Visualization system for clinical internal medicine and visualization method thereof

Abstract

The invention relates to the technical field of clinical medical data visualization, in particular to a visualization system for clinical internal medicine and a visualization method thereof. The method comprises the following steps: acquiring basic information of a clinical patient; according to the basic information of the clinical patient, performing medical treatment and medication event analysis, and performing triple entity relationship processing to generate patient medical record relationship map data; performing patient drug tolerance analysis based on the patient medical record relation map data to generate patient drug tolerance characteristic data; performing clinical physiological index risk probabilistic reasoning according to the drug tolerance characteristic data of the patient to generate individual clinical medication risk data; according to the individual clinical medication risk data, individual characteristic medicine dosage adjustment is carried out, medicine dosage curve visual dynamic display is carried out, and visual clinical medication response data is generated. According to the method, personalized and visual medication decision support is provided for clinicians through patient medication conditions and real-time physiological index analysis.

Description

Technical Field

[0001] The present invention relates to the technical field of clinical medical data visualization, and in particular to a visualization system for clinical internal medicine and a visualization method thereof. Background Art

[0002] As an important branch of the medical field, clinical internal medicine covers a wide range of disease diagnosis and treatment, among which drug therapy is the core means of internal medicine disease management. The effectiveness and safety of drug therapy are directly related to the patient's health and quality of life. Therefore, it is crucial to formulate a reasonable medication plan. Doctors need to consider many factors when formulating medication plans, including the patient's condition, physiological state, pharmacological effects of drugs, drug interactions, etc. The amount of information is huge and complex, and it is difficult to fully and accurately analyze and judge based on the doctor's experience and memory alone. In the traditional clinical medicine medication model, doctors often prescribe according to the dosage range recommended by clinical guidelines. These guidelines are usually based on the average value or range obtained from large-scale population studies. Ignoring individual differences between patients, resulting in uneven medication effects and even serious adverse reactions, which bring potential risks to patients' health and cannot be visualized to assist doctors in making relevant decisions. Summary of the invention

[0003] Based on this, the present invention provides a visualization system for clinical internal medicine and a visualization method thereof to solve at least one of the above technical problems.

[0004] To achieve the above object, a clinical internal medicine visualization method comprises the following steps:

[0005] Step S1: Obtain basic information of clinical patients; segment medical record terminology text according to the basic information of clinical patients to generate medical record text terminology segmentation data; perform medical treatment and medication event analysis on the medical record text terminology segmentation data to generate time series medical record medication event data;

[0006] Step S2: performing medical entity recognition on the medical record text term segmentation data to obtain medical entity data; performing medical term relationship mapping based on the medical entity data to generate medical term entity mapping relationship data; performing triple entity relationship processing based on the medical term entity mapping relationship data to generate patient medical record relationship graph data;

[0007] Step S3: Based on the patient medical record relationship graph data, Bayesian network node processing is performed to generate Bayesian network structure data; based on the time series medical record medication event data, the patient's drug tolerance analysis is performed to generate the patient's drug tolerance characteristic data; based on the patient's drug tolerance characteristic data, the patient's clinical physiological indicators are dynamically monitored to generate dynamic patient clinical physiological indicator data; based on the dynamic patient clinical physiological indicator data, the patient's drug tolerance characteristic data is used to perform clinical physiological indicator risk probability reasoning on the Bayesian network structure data to generate individual clinical medication risk data;

[0008] Step S4: Evaluate the patient's metabolic capacity based on individual clinical medication risk data and generate revised drug metabolic clearance baseline data;

[0009] Step S5: Acquire clinical physiological effect drug data; adjust the individual characteristic drug dosage of the clinical physiological effect drug data by correcting the drug metabolism clearance baseline data, and visualize the drug dosage curve dynamically to generate visualized clinical medication response data.

[0010] The present invention constructs a patient medical record relationship map by segmenting, entity recognition and relationship mapping of the patient's medical record text, so that the system can fully and deeply understand the patient's condition and medical history, avoiding the omission and deviation of information caused by judging only by the doctor's experience and memory. By segmenting, entity recognition and relationship mapping of the patient's medical record text, and constructing a patient medical record relationship map, the system can fully and deeply understand the patient's condition and medical history, avoiding the omission and deviation of information caused by judging only by the doctor's experience and memory. Through the steps of automating the processing of patient information, constructing a medical record relationship map, and performing risk probability reasoning, the medication plan formulation process is greatly simplified, saving the doctor's time and energy, enabling the doctor to focus more on the diagnosis and treatment of the patient's condition, and improving the efficiency of diagnosis and treatment. By converting complex medical data into intuitive visualization charts, such as drug dosage curves, the system can clearly display the basis for the formulation of the medication plan and the expected effect. Therefore, a visualization method for clinical internal medicine of the present invention extracts key medical treatment and medication event data by segmenting and analyzing the medical record text, constructs a patient medical record relationship map, and reveals the complex relationship between the patient's condition and medication. On this basis, the Bayesian network is used to dynamically monitor the patient's drug tolerance and clinical physiological indicators and conduct risk probability inference, accurately adjust the drug dosage, and dynamically display the drug dosage curve through visualization means. This not only improves the scientificity and accuracy of the medication plan, but also provides clinicians with an efficient and convenient visualization tool, thereby significantly improving the effectiveness and safety of drug treatment.

[0011] Preferably, step S1 comprises the following steps:

[0012] Step S11: Obtaining basic clinical patient information;

[0013] Step S12: extracting historical medical records based on the basic information of clinical patients to generate historical medical record data of clinical patients;

[0014] Step S13: performing historical data cleaning processing according to the historical medical record data of clinical patients, and performing terminology text segmentation on the electronic medical record standard data through a preset medical term dictionary to generate medical record text term segmentation data;

[0015] Step S14: Analyze the medical record text term segmentation data for medical treatment and medication events, and perform time axis alignment processing to generate time-series medical record medication event data.

[0016] The present invention effectively improves the quality and reliability of data through the cleaning process of historical data. Clinical medical record data usually has problems such as inconsistent formats and non-standardized content, and direct use will affect the accuracy of subsequent analysis. The process of data cleaning can remove noise data, unify data formats, correct erroneous information, ensure the purity of the data basis for subsequent analysis, and ensure the reliability of the final analysis results. Furthermore, the electronic medical record standard data is segmented into terminology texts through a preset medical terminology dictionary, and the unstructured medical record text is converted into a structured data form, laying the foundation for subsequent computer processing. Timeline alignment processing can arrange medication events in different time periods in chronological order, making it convenient for the system to track the patient's condition changes and medication effects, and providing an important reference basis for the formulation of individualized medication plans.

[0017] Preferably, step S13 comprises the following steps:

[0018] Step S131: Perform text encoding format detection on clinical patient historical medical record data to obtain medical record encoding data;

[0019] Step S132: decoding the historical medical record data of clinical patients through the medical record coding data to generate decoded medical record text data;

[0020] Step S133: using a preset regular expression to match special symbols in the decoded medical record text data, and removing non-medical text to obtain simplified historical medical record text data;

[0021] Step S134: matching the separator positions of the simplified historical medical record text data with the preset medical paragraph separators, and performing separator position analysis to generate separator position data;

[0022] Step S135: segmenting the simplified historical medical record text data into text segments according to the separator position data to obtain historical medical record segmentation data;

[0023] Step S136: Use a preset medical term dictionary to perform paragraph type keyword matching on the historical medical record paragraph segmentation data to obtain paragraph type label data;

[0024] Step S137: Utilize the paragraph type label data to merge the same type of paragraphs in the simplified historical medical record text data to generate medical record text term segmentation data.

[0025] The present invention detects and decodes text encoding formats, and the system can uniformly convert medical record data of different formats into a processable text format, eliminating the obstacles caused by data format differences. Special symbol matching and non-medical text elimination further purify the data, remove irrelevant information, such as punctuation marks, headers and footers, etc., retain key medical text content, and improve the efficiency and accuracy of data analysis. Using preset medical paragraph delimiters and delimiter position analysis, the system can divide the medical record text into different paragraphs, and match each paragraph with type keywords according to the medical term dictionary, marking the paragraph type, such as diagnosis, medication, examination, etc. This structured processing method enables the system to organize and manage medical record information more effectively, facilitating subsequent analysis and retrieval. By merging paragraphs of the same type, the system integrates similar information scattered in different locations, such as merging the same diagnostic information mentioned multiple times into one, avoiding redundancy and duplication of information, and further improving the efficiency of data processing.

[0026] Preferably, step S2 comprises the following steps:

[0027] Step S21: performing medical entity recognition on the medical record text term segmentation data to obtain medical entity data;

[0028] Step S22: performing entity disambiguation on the medical entity data, and using a pre-trained medical relationship extraction model to extract the relationship between entities to obtain dependency tree data;

[0029] Step S23: performing entity direct relationship analysis according to the dependency tree data to obtain entity direct relationship data;

[0030] Step S24: classifying the relationships according to the entity direct relationship data, and performing potential entity relationship reasoning to generate potential relationship data;

[0031] Step S25: performing weighted confidence evaluation on entity direct relationship data and potential relationship data, and performing medical term relationship mapping to generate medical term entity mapping relationship data;

[0032] Step S26: Perform triple entity relationship processing according to the medical term entity mapping relationship data to generate patient medical record relationship graph data.

[0033] The present invention extracts key medical entities, such as diseases, symptoms, and drugs, from medical record texts, laying the foundation for subsequent relationship extraction. Entity disambiguation solves the ambiguity problem of entities with the same name, and the accuracy of entities can be ensured through disambiguation. Using pre-trained medical relationship extraction models and dependency tree analysis, the system can identify direct relationships between entities, such as "the patient suffers from hypertension", thereby preliminarily constructing an entity relationship network. Not all relationships in medical record texts are explicitly expressed, and some relationships need to be obtained through reasoning. Therefore, the system uses potential entity relationship reasoning to dig out potential relationships hidden behind the text, such as inferring the patient's disease based on the patient's symptoms and test results, so as to more comprehensively understand the patient's condition. Weighted confidence assessment of direct and potential relationships can distinguish the importance and credibility of relationships, and avoid interference of incorrect relationships in subsequent analysis.

[0034] Preferably, step S3 comprises the following steps:

[0035] Step S31: performing Bayesian network node processing based on the patient medical record relationship graph data to generate Bayesian network structure data;

[0036] Step S32: Deeply mine the patient's medication history through the time-series medical record medication event data to generate patient medication history feature data;

[0037] Step S33: performing a drug tolerance analysis on the patient according to the characteristic data of the patient's medication history, and generating the patient's drug tolerance characteristic data;

[0038] Step S34: using the patient's drug tolerance characteristic data to perform inter-node conditional probability processing on the Bayesian network structure data to generate Bayesian node conditional parameters;

[0039] Step S35: dynamically monitoring the patient's clinical physiological indicators according to the patient's drug tolerance characteristic data, and generating dynamic patient clinical physiological indicator data;

[0040] Step S36: Perform clinical physiological indicator risk probability reasoning on the dynamic patient clinical physiological indicator data through Bayesian node condition parameters to generate individual clinical medication risk data.

[0041] The present invention constructs a Bayesian network structure based on the patient medical record relationship map, takes medical entities such as diseases, symptoms, and drugs as network nodes, and defines the connections between nodes based on the relationships between entities, thereby establishing a probability model that can reflect the patient's condition and medication situation. By deeply mining the time-series medical record medication event data, the system extracts the patient's medication history characteristics, such as medication type, dosage, duration, etc., and conducts drug tolerance analysis on this basis to evaluate the patient's sensitivity and tolerance to different drugs. This information is used to correct the conditional probability between nodes of the Bayesian network, so that the model is more in line with the individual characteristics of the patient.

[0042] Preferably, step S33 includes the following steps:

[0043] Step S331: classify the drugs according to the patient's medication history characteristic data to obtain the patient's classified medication data;

[0044] Step S332: analyzing the drug dosage change of the patient's classified medication data to obtain the patient's medication dosage change data;

[0045] Step S333: Evaluate the efficacy index value of the patient's medication history characteristic data through the patient's classified medication data to generate the patient's medication efficacy index value;

[0046] Step S334: using the patient's medication dosage change data to process the patient's medication efficacy index value into a therapeutic effect response curve, and performing a drug sensitivity analysis to obtain drug sensitivity assessment data;

[0047] Step S335: Calculate the dose adjustment factor required for the target therapeutic effect according to the drug sensitivity assessment data, and perform drug tolerance numerical label processing to obtain numerical tolerance data;

[0048] Step S336: Integrate the numerical tolerance data, the patient's medication dosage change data, and the patient's medication efficacy index value to generate the patient's medication tolerance characteristic data.

[0049] The present invention classifies drugs according to the patient's medication history and analyzes the dosage changes of different categories of drugs, which helps to distinguish the differences in patients' responses to different types of drugs. At the same time, the system evaluates the efficacy index values ​​of patients after medication, such as changes in indicators such as blood pressure and blood sugar, and combines these indicators with the dosage change data to process the efficacy response curve, thereby more intuitively showing the relationship between drug efficacy and dosage. By performing drug sensitivity analysis on the efficacy response curve, the system can evaluate the patient's sensitivity to different drugs, for example, some patients react more strongly or more slowly to a certain drug than other patients. Based on the drug sensitivity assessment data, the system calculates the dosage adjustment factor required for the target efficacy, that is, how much adjustment needs to be made to the standard dose in order to achieve the expected therapeutic effect.

[0050] Preferably, step S35 includes the following steps:

[0051] Step S351: Processing a key monitoring organ list according to the patient's drug tolerance characteristic data to generate key monitoring organ data;

[0052] Step S352: configuring monitoring indicators for key monitoring organ data through a preset list of monitoring organ physiological indicators to generate monitoring indicator configuration data;

[0053] Step S353: monitoring the patient's organ physiological indexes using clinical monitoring equipment based on the monitoring index configuration data to generate clinical organ monitoring physiological index data;

[0054] Step S354: performing organ function status classification on the clinical organ monitoring physiological index data to obtain organ function status classification data;

[0055] Step S355: Based on the organ function status classification data, the physiological index change rate of the clinical organ monitoring physiological index data is processed to obtain dynamic patient clinical physiological index data.

[0056] The present invention determines the organs that need to be monitored based on the patient's drug tolerance characteristic data. For example, if the patient's liver tolerance to a certain drug is poor, it is necessary to focus on monitoring liver function indicators. This monitoring scheme based on individual characteristics is more targeted than conventional comprehensive monitoring, can effectively improve monitoring efficiency, and reduce unnecessary inspections. The system configures specific physiological indicators that need to be monitored, such as alanine aminotransferase and aspartate aminotransferase in liver function indicators, based on a preset list of monitored organ physiological indicators and key monitored organ data. The patient's organ physiological indicator data is collected in real time through clinical monitoring equipment, and the data is graded for organ function status, for example, liver function status is divided into normal, mildly abnormal, moderately abnormal, and severely abnormal, etc., so as to more intuitively reflect the health status of the patient's organ function.

[0057] Preferably, step S4 comprises the following steps:

[0058] Step S41: Evaluate the metabolic capacity of the patient according to the individual clinical medication risk data to obtain metabolic capacity score data of the patient;

[0059] Step S42: Calculating the initial adjusted drug clearance rate based on the patient's metabolic capacity score data to generate initial drug adjusted drug clearance rate data;

[0060] Step S43: Acquire the patient's renal function index data and the patient's liver function index data;

[0061] Step S44: Based on the patient's renal function index data and the patient's liver function index data, the initial drug-adjusted clearance rate data is corrected for liver and kidney baseline values ​​to generate corrected drug metabolism clearance baseline data.

[0062] The present invention evaluates the overall metabolic capacity of the patient based on individual clinical medication risk data, and quantifies it into a metabolic capacity score. The score comprehensively considers the patient's various physiological indicators, disease conditions, medication history and other factors, and more comprehensively reflects the patient's metabolic capacity level. Based on the patient's metabolic capacity score, the system calculates the initial adjustment clearance rate of the drug, which is the starting reference value for drug dosage adjustment. Since drug metabolism and clearance mainly depend on the function of the liver and kidneys, the system obtains the patient's renal function index data and liver function index data, such as creatinine clearance, alanine aminotransferase, etc. Using these index data, the system corrects the initial drug adjustment clearance rate with liver and kidney baseline values ​​to generate more accurate corrected drug metabolism and clearance baseline data. Individual differences in liver and kidney function of patients are taken into account. For example, for patients with impaired liver and kidney function, their drug metabolism and clearance capacity will be reduced, so it is necessary to reduce the drug dose or extend the dosing interval accordingly. By accurately correcting the drug metabolism and clearance baseline, the system can provide a more reliable basis for subsequent individualized drug dosage adjustments, thereby improving the effectiveness and safety of medication, and avoiding drug accumulation or insufficient efficacy caused by differences in drug metabolism and clearance capacity.

[0063] Preferably, step S5 comprises the following steps:

[0064] Step S51: querying the clinical disease-effect drugs of the patient through a preset drug database to generate clinical physiological effect drug data;

[0065] Step S52: adjusting the individual characteristic drug dosage of the clinical physiological effect drug data by correcting the drug metabolism clearance baseline data to obtain personalized dosage range data;

[0066] Step S53: Based on the personalized dose range data, the patient's drug dose response is simulated through the clinical physiological drug data, and the response index is predicted to obtain the predicted drug concentration data, the predicted efficacy index value and the adverse reaction probability data at different doses;

[0067] Step S54: performing drug dose curve axis processing on the predicted drug concentration data, predicted efficacy index value and adverse reaction probability data through the personalized dose range data to generate dose response curve data;

[0068] Step S55: Perform interactive control processing according to the dose response curve data, and perform dynamic visualization of the curve to generate visualized clinical medication response data.

[0069] In order to predict the medication effect under different doses, the present invention systematically simulates the medication response of drug doses, predicts drug concentrations, efficacy index values ​​and the probability of adverse reactions. These prediction results provide a scientific basis for doctors to select the best dose, which helps to balance efficacy and safety. The drug dose curve axis is processed by personalized dose range data to predict drug concentration data, predict efficacy index values ​​and adverse reaction probability data, and dose response curve data is generated. This curve data can intuitively display the relationship between drug dose and drug concentration, efficacy index and adverse reaction probability, so that doctors can clearly understand the changing trend of drugs under different doses, which helps to select the dose with the best efficacy and the lowest risk. Finally, interactive control processing is performed according to the dose response curve data, and the curve is visualized and dynamically displayed to generate visualized clinical medication response data. This visualization method not only makes the complex dose response relationship intuitive and easy to understand, but also provides an interactive operation interface. Doctors can adjust the dose and observe the changes in the dose response curve in real time, and more intuitively understand the efficacy and risk of the drug, so as to make more accurate medication decisions.

[0070] The present invention also provides a clinical internal medicine visualization system, which executes the clinical internal medicine visualization method as described above, and the clinical internal medicine visualization system comprises:

[0071] The medical record data preprocessing module is used to obtain the basic information of clinical patients; segment the medical record terminology text according to the basic information of clinical patients to generate medical record text terminology segmentation data; analyze the medical treatment and medication events of the medical record text terminology segmentation data to generate time series medical record medication event data;

[0072] The entity relationship graph module is used to perform medical entity recognition on the medical record text term segmentation data to obtain medical entity data; perform medical term relationship mapping based on the medical entity data to generate medical term entity mapping relationship data; perform triple entity relationship processing based on the medical term entity mapping relationship data to generate patient medical record relationship graph data;

[0073] The tolerance index risk module is used to process Bayesian network nodes based on the patient medical record relationship map data to generate Bayesian network structure data; perform patient drug tolerance analysis based on the time-series medical record medication event data to generate patient drug tolerance characteristic data; dynamically monitor the patient's clinical physiological indicators based on the patient's drug tolerance characteristic data to generate dynamic patient clinical physiological indicator data; perform clinical physiological indicator risk probability reasoning on the Bayesian network structure data based on the dynamic patient clinical physiological indicator data using the patient's drug tolerance characteristic data to generate individual clinical medication risk data;

[0074] The metabolic capacity assessment module is used to assess the metabolic capacity of patients based on individual clinical medication risk data and generate modified drug metabolic clearance baseline data;

[0075] The dose adjustment visualization module is used to obtain clinical physiological drug data; by correcting the drug metabolism clearance baseline data, the clinical physiological drug data is adjusted for individual characteristic drug doses, and the drug dose curve is visualized and dynamically displayed to generate visualized clinical medication response data. BRIEF DESCRIPTION OF THE DRAWINGS

[0076] Figure 1 A schematic diagram of the steps of a visualization method for clinical internal medicine of the present invention;

[0077] Figure 2 for Figure 1 Detailed implementation steps of step S2 in the flowchart;

[0078] Figure 3 for Figure 1 Detailed implementation steps of step S4 in FIG.

[0079] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0080] The technical method of the present invention is described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by technicians in this field without creative work are within the scope of protection of the present invention.

[0081] In addition, the accompanying drawings are only schematic illustrations of the present invention and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. The functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor methods and / or microcontroller methods.

[0082] It should be understood that, although the terms "first", "second", etc. may be used herein to describe various units, these units should not be limited by these terms. These terms are used only to distinguish one unit from another unit. For example, without departing from the scope of the exemplary embodiments, the first unit may be referred to as the second unit, and similarly the second unit may be referred to as the first unit. The term "and / or" used herein includes any and all combinations of one or more of the listed associated items.

[0083] To achieve this, please refer to Figures 1 to 3 The present invention provides a visualization method for clinical internal medicine, comprising the following steps:

[0084] Step S1: Obtain basic information of clinical patients; segment medical record terminology text according to the basic information of clinical patients to generate medical record text terminology segmentation data; perform medical treatment and medication event analysis on the medical record text terminology segmentation data to generate time series medical record medication event data;

[0085] Step S2: performing medical entity recognition on the medical record text term segmentation data to obtain medical entity data; performing medical term relationship mapping based on the medical entity data to generate medical term entity mapping relationship data; performing triple entity relationship processing based on the medical term entity mapping relationship data to generate patient medical record relationship graph data;

[0086] Step S3: Based on the patient medical record relationship graph data, Bayesian network node processing is performed to generate Bayesian network structure data; based on the time series medical record medication event data, the patient's drug tolerance analysis is performed to generate the patient's drug tolerance characteristic data; based on the patient's drug tolerance characteristic data, the patient's clinical physiological indicators are dynamically monitored to generate dynamic patient clinical physiological indicator data; based on the dynamic patient clinical physiological indicator data, the patient's drug tolerance characteristic data is used to perform clinical physiological indicator risk probability reasoning on the Bayesian network structure data to generate individual clinical medication risk data;

[0087] Step S4: Evaluate the patient's metabolic capacity based on individual clinical medication risk data and generate revised drug metabolic clearance baseline data;

[0088] Step S5: Acquire clinical physiological effect drug data; adjust the individual characteristic drug dosage of the clinical physiological effect drug data by correcting the drug metabolism clearance baseline data, and visualize the drug dosage curve dynamically to generate visualized clinical medication response data.

[0089] In an embodiment of the present invention, the clinical internal medicine visualization method comprises the following steps:

[0090] Step S1: Obtain basic information of clinical patients; segment medical record terminology text according to the basic information of clinical patients to generate medical record text terminology segmentation data; perform medical treatment and medication event analysis on the medical record text terminology segmentation data to generate time series medical record medication event data;

[0091] In an embodiment of the present invention, the basic information of clinical patients, including patient ID, name, gender, age, past medical history, allergy history, etc., is obtained through a hospital information system (HIS) or an electronic medical record system (EMR). The complete electronic medical record text of the patient during hospitalization is obtained, for example: "Patient's chief complaint: cough and sputum for 3 days, accompanied by fever. Past history: history of hypertension for 5 years, regular use of nifedipine to control blood pressure. Physical examination: body temperature 38.5°C, coarse breath sounds, and moist rales can be heard. Auxiliary examination: blood routine shows elevated white blood cell count. Diagnosis: acute bronchitis. Doctor's order: give oral cefixime, twice a day, 0.2g each time; oral acetaminophen, three times a day, 0.5g each time." The medical record term text is segmented according to the basic information of clinical patients. Using natural language processing technology (NLP), such as regular expressions and syntactic analysis, the medical record text is divided into sections such as chief complaint, current medical history, past medical history, physical examination, auxiliary examination, diagnosis, and doctor's advice to generate structured medical record text term segmentation data. For example, the above medical record text is divided into: Chief complaint: "cough and sputum for 3 days, accompanied by fever"; Past medical history: "History of hypertension for 5 years, regular use of nifedipine to control blood pressure"; and so on. Medical treatment and medication event analysis is performed on the medical record text term segmentation data, especially the doctor's advice and medication record parts. Extract medication information during each visit / hospitalization, including drug name, dosage, usage, medication time, etc. Arrange these medication information in chronological order to generate time-series medical record medication event data. For example, based on the above medical record text, two medication events, "Cefixime, 0.2 g, oral, twice a day" and "Acetaminophen, 0.5 g, oral, three times a day", can be extracted, and their start time of medication can be recorded.

[0092] Step S2: performing medical entity recognition on the medical record text term segmentation data to obtain medical entity data; performing medical term relationship mapping based on the medical entity data to generate medical term entity mapping relationship data; performing triple entity relationship processing based on the medical term entity mapping relationship data to generate patient medical record relationship graph data;

[0093] In an embodiment of the present invention, medical named entity recognition (NER) technology, such as a BiLSTM-CRF model, is used to perform medical entity recognition on the medical record text term segmentation data generated in step S1. Medical entities such as diseases, symptoms, signs, drugs, and examination indicators are identified. For example, three symptom entities of "cough", "sputum" and "fever" are identified from "cough, sputum for 3 days, accompanied by fever". The identified medical entities are mapped with standard medical terms in the medical knowledge graph to generate medical term entity mapping relationship data. For example, "cough" is mapped to the "cough" concept in the medical knowledge graph, and "cefixime" is mapped to the "cefixime" drug concept. Based on the medical term entity mapping relationship data, a patient medical record relationship graph is constructed. Medical entities are used as nodes, and the relationships between entities are used as edges to form a graph structure. For example, you can construct relationship triplets such as "acute bronchitis"-cause-"cough", "acute bronchitis"-treatment-"cefixime", and add these triplets to the patient's medical record relationship graph.

[0094] Step S3: Based on the patient medical record relationship graph data, Bayesian network node processing is performed to generate Bayesian network structure data; based on the time series medical record medication event data, the patient's drug tolerance analysis is performed to generate the patient's drug tolerance characteristic data; based on the patient's drug tolerance characteristic data, the patient's clinical physiological indicators are dynamically monitored to generate dynamic patient clinical physiological indicator data; based on the dynamic patient clinical physiological indicator data, the patient's drug tolerance characteristic data is used to perform clinical physiological indicator risk probability reasoning on the Bayesian network structure data to generate individual clinical medication risk data;

[0095] In an embodiment of the present invention, a Bayesian network structure is constructed using patient medical record relationship graph data. The nodes in the graph are used as nodes of the Bayesian network, the relationships between the nodes are used as directed edges of the Bayesian network, and the conditional probabilities between the nodes are learned based on medical knowledge or data. For example, "acute bronchitis" can be used as a parent node, and "cough" can be used as a child node, and there is a conditional probability relationship between the two. According to the time series medical record medication event data generated in step S1, the patient's tolerance to different drugs is analyzed. For example, by analyzing the occurrence of adverse reactions and changes in efficacy after a patient takes a certain drug, the patient's tolerance to the drug is judged. The patient's tolerance information to different drugs is converted into feature data, for example, the degree of tolerance can be quantified as a value between 0 and 1. The patient's clinical physiological indicators are dynamically monitored, such as blood pressure, heart rate, body temperature, liver and kidney function, etc. The monitored physiological indicator data is recorded in chronological order to generate dynamic patient clinical physiological indicator data. Combined with the dynamic patient clinical physiological indicator data and the patient's drug tolerance feature data, the Bayesian network is used to perform clinical physiological indicator risk probability reasoning. For example, based on the patient's medication status and drug tolerance characteristics, the probability of the patient's liver and kidney function damage can be inferred, and the inferred risk probability is used as individual clinical medication risk data.

[0096] Step S4: Evaluate the patient's metabolic capacity based on individual clinical medication risk data and generate revised drug metabolic clearance baseline data;

[0097] In an embodiment of the present invention, a risk assessment model is established to map individual clinical medication risk data to the patient's metabolic capacity level. The model can be constructed based on expert experience, clinical data or machine learning algorithms. For example, a model based on Logistic regression can be constructed, with various risk probabilities (such as liver damage risk, renal damage risk) calculated in step S3 as input features, and the probability of impaired metabolic capacity of the patient is output. Assume that in the model training data, the probability of impaired metabolic capacity of patients with a liver damage risk probability greater than 0.8 and a renal damage risk probability greater than 0.6 is close to 1. Then when the liver damage risk probability of a new patient is 0.85 and the renal damage risk probability is 0.7, the model can predict that the patient has a higher probability of impaired metabolic capacity. A standard drug metabolism clearance baseline needs to be defined. The baseline can be set based on the pharmacokinetic parameters of a healthy population, for example, the average clearance value provided in the drug instructions can be referred to. Assume that the average clearance rate of a certain drug in a healthy population is 50 mL / min. A standard drug metabolism clearance baseline needs to be defined. The baseline can be set based on the pharmacokinetic parameters of a healthy population, for example, the average clearance value provided in the drug instructions can be referred to. Assume that the average clearance of a drug in healthy people is 50 mL / min. Correct the drug metabolism clearance baseline based on the patient's metabolic capacity assessment results. The correction method can be determined based on the specific pharmacokinetic model. For example, the standard clearance value can be reduced proportionally based on the probability of impaired metabolic capacity of the patient. Assuming that the probability of impaired metabolic capacity of the patient is 0.8, the standard clearance rate of 50 mL / min can be reduced by 40% (0.8×50 mL / min=40 mL / min), resulting in a corrected clearance rate of 10 mL / min.

[0098] Step S5: Acquire clinical physiological effect drug data; adjust the individual characteristic drug dosage of the clinical physiological effect drug data by correcting the drug metabolism clearance baseline data, and visualize the drug dosage curve dynamically to generate visualized clinical medication response data.

[0099] In an embodiment of the present invention, clinical physiological effect drug data, including pharmacological effects, metabolic kinetic parameters, etc. of the drug, are obtained from a drug database or pharmacopoeia. According to the corrected drug metabolic clearance baseline data generated in step S4, the clinical physiological effect drug data is adjusted for individual characteristic drug dosage. For example, according to the drug clearance rate of the patient, the dosage and dosing interval of the drug are adjusted to achieve the best therapeutic effect and reduce the occurrence of adverse reactions. The adjusted drug dosage information is visualized and dynamically displayed, for example, a curve chart showing the change of drug concentration over time is drawn. The visualized drug dosage information and the clinical physiological index data of the patient are integrated together to generate visualized clinical medication response data, providing intuitive medication guidance for clinicians. For example, the drug concentration curve and the patient's blood pressure, heart rate and other index curves can be displayed in the same chart, which is convenient for doctors to observe the efficacy and safety of the drug.

[0100] Preferably, step S1 comprises the following steps:

[0101] Step S11: Obtaining basic clinical patient information;

[0102] Step S12: extracting historical medical records based on the basic information of clinical patients to generate historical medical record data of clinical patients;

[0103] Step S13: performing historical data cleaning processing according to the historical medical record data of clinical patients, and performing terminology text segmentation on the electronic medical record standard data through a preset medical term dictionary to generate medical record text term segmentation data;

[0104] Step S14: Analyze the medical record text term segmentation data for medical treatment and medication events, and perform time axis alignment processing to generate time-series medical record medication event data.

[0105] In an embodiment of the present invention, basic information of clinical patients is obtained through a hospital information system (HIS) or an electronic medical record system (EMR). This information includes, but is not limited to, patient ID, name, gender, age, height, weight, contact information, allergy history, family history, etc. For example, basic information such as patient ID 123456, name Zhang San, gender male, and age 50 years old can be obtained from the HIS system. According to the patient ID obtained in step S11, the historical medical record data of the patient is extracted from the electronic medical record system. The historical medical record data may include records of the patient's previous visits or hospitalizations, such as outpatient medical records, inpatient medical records, examination reports, test results, etc. For example, all historical medical records of patients with patient ID 123456 are extracted, and the historical medical record data extracted in step S12 are cleaned. The cleaning process includes removing duplicate data, correcting erroneous data, processing missing data, etc. For example, if the same patient undergoes the same examination twice on the same day, one of the duplicate data can be removed. The cleaned data is more standardized and accurate, which is conducive to subsequent analysis. In addition, the cleaned electronic medical record standard data is segmented into terminology text using a preset medical term dictionary. For example, the medical record text is divided into sections such as the chief complaint, current medical history, past medical history, physical examination, auxiliary examination, diagnosis, and doctor's advice. For example, "the patient's chief complaint: cough, sputum for 3 days, accompanied by fever" is divided into the "chief complaint" section, and "past medical history: 5 years of hypertension history" is divided into the "past medical history" section. The medical record text term segmentation data generated in step S13, especially the doctor's advice and medication record parts, are analyzed for medical treatment and medication events. Medication information is extracted during each visit / hospitalization, including drug name (e.g., "amoxicillin", "acetaminophen"), dosage (e.g., "250 mg", "500 mg"), usage (e.g., "oral", "intravenous injection"), medication frequency (e.g., "three times a day", "once every 8 hours"), and medication start time. For example, from the doctor's order "Amoxicillin 250 mg orally, three times a day", the drug name "Amoxicillin", dosage "250 mg", usage "oral", and frequency of use "three times a day" are extracted. The extracted medication information is arranged in chronological order and time axis alignment is performed to generate time series medical record medication event data.

[0106] Preferably, step S13 comprises the following steps:

[0107] Step S131: Perform text encoding format detection on clinical patient historical medical record data to obtain medical record encoding data;

[0108] Step S132: decoding the historical medical record data of clinical patients through the medical record coding data to generate decoded medical record text data;

[0109] Step S133: using a preset regular expression to match special symbols in the decoded medical record text data, and removing non-medical text to obtain simplified historical medical record text data;

[0110] Step S134: matching the separator positions of the simplified historical medical record text data with the preset medical paragraph separators, and performing separator position analysis to generate separator position data;

[0111] Step S135: segmenting the simplified historical medical record text data into text segments according to the separator position data to obtain historical medical record segmentation data;

[0112] Step S136: Use a preset medical term dictionary to perform paragraph type keyword matching on the historical medical record paragraph segmentation data to obtain paragraph type label data;

[0113] Step S137: Utilize the paragraph type label data to merge the same type of paragraphs in the simplified historical medical record text data to generate medical record text term segmentation data.

[0114] In an embodiment of the present invention, the clinical patient historical medical record data extracted from step S12 is obtained. Then, the encoding format of the medical record data is detected using a character encoding detection library (e.g., a chardet library). Common encoding formats include UTF-8, GBK, GB2312, Latin-1, etc. For example, it is detected that a certain medical record data is encoded in UTF-8. The detected encoding format information is recorded as medical record encoding data. According to the medical record encoding data obtained in step S131, the historical medical record data is decoded using a corresponding decoding method. For example, if the medical record encoding data is UTF-8, the medical record data is decoded using a UTF-8 decoder. The decoded data is readable text data, for example: "Patient's chief complaint: Headache for 3 days". The decoded text data is saved as decoded medical record text data. A preset regular expression is used to match and eliminate non-medical text and special symbols, such as advertising information, contact information, special characters, etc. Preset medical paragraph separators, such as "Chief complaint:", "Current medical history:", "Past medical history:", "Physical examination:", "Diagnosis:", etc. Use the string matching function to find the positions where these delimiters appear in the simplified historical medical record text data. For example, in the text "Patient's chief complaint: headache for 3 days. Current medical history: cold for one week", "chief complaint:" appears at the 3rd character position, and "current medical history:" appears at the 13th character position. Record these position information as separator position data. According to the separator position data obtained in step S134, the simplified historical medical record text data is divided into different paragraphs. For example, according to the separator position data in the above example, the text can be divided into two paragraphs: "Patient's chief complaint: headache for 3 days." and "Current medical history: cold for one week". Save the segmented paragraph data as historical medical record paragraph segmentation data. The preset medical term dictionary contains various medical keywords and their corresponding paragraph types. For example, the "chief complaint" keyword corresponds to the "chief complaint" paragraph type, and the "current medical history" keyword corresponds to the "current medical history" paragraph type. Traverse the historical medical record paragraph segmentation data and match each paragraph with the keywords in the dictionary. For example, the paragraph "Patient's chief complaint: headache for 3 days." can match the keyword "chief complaint", so it is marked as the "chief complaint" paragraph type. The type labels of all paragraphs are recorded as paragraph type label data. According to the paragraph type label data obtained in step S136, the paragraphs of the same type in the simplified historical medical record text data are merged. For example, if there are multiple "chief complaint" paragraphs, they are merged into one "chief complaint" paragraph. For example, "Chief complaint: headache for 3 days." and "Chief complaint: accompanied by nausea" are merged into "Chief complaint: headache for 3 days, accompanied by nausea". The merged data contains paragraphs of different types, such as chief complaint, current medical history, past history, etc., forming the final medical record text term segmentation data.

[0115] As an example of the present invention, refer to Figure 2 As shown, Figure 1The detailed implementation step flow chart of step S2 in the embodiment includes:

[0116] Step S21: performing medical entity recognition on the medical record text term segmentation data to obtain medical entity data;

[0117] In an embodiment of the present invention, a pre-trained medical named entity recognition (NER) model, such as a model based on the BERT architecture, is used to process the medical record text term segmentation data output by step S1. The model identifies medical entities appearing in the text, such as diseases (such as "hypertension", "diabetes"), symptoms (such as "headache", "cough"), drugs (such as "aspirin", "nifedipine"), examination items (such as "blood routine", "CT scan"), etc. For example, for the input text "Patient's chief complaint: cough, sputum for 3 days, accompanied by fever", the model can identify three symptom entities of "cough", "sputum", and "fever". The identified entities and their location information in the text are recorded to form medical entity data.

[0118] Step S22: performing entity disambiguation on the medical entity data, and using a pre-trained medical relationship extraction model to extract the relationship between entities to obtain dependency tree data;

[0119] In an embodiment of the present invention, since the same medical term refers to different concepts in different contexts, it is necessary to disambiguate the identified medical entities. For example, "hypertension" can refer to primary hypertension or secondary hypertension. The specific meaning of the entity can be determined using a medical knowledge graph or a context-based disambiguation algorithm. For example, if the description of "renal artery stenosis" also appears in the medical record, it can be inferred that "hypertension" refers to secondary hypertension. After completing the entity disambiguation, a pre-trained medical relationship extraction model, such as a model based on a graph convolutional network (GCN), is used to analyze the relationship between entities in the text. The model constructs a dependency tree to represent the dependency relationship between words in a sentence. For example, in the sentence "The patient took nifedipine due to hypertension", there is a "reason for taking" relationship between "hypertension" and "nifedipine". The extracted entity relationship and dependency tree structure are saved as dependency tree data.

[0120] Step S23: performing entity direct relationship analysis according to the dependency tree data to obtain entity direct relationship data;

[0121] In an embodiment of the present invention, the dependency tree data generated in step S22 is analyzed to extract direct relationships between entities. A direct relationship refers to a relationship that directly connects two entities in a dependency tree. For example, in the dependency tree corresponding to the sentence "The patient takes nifedipine due to hypertension", "hypertension" and "nifedipine" are directly connected, and the relationship between them is "reason for taking". The extracted direct relationship, such as ("hypertension", "reason for taking", "nifedipine"), is saved as entity direct relationship data.

[0122] Step S24: classifying the relationships according to the entity direct relationship data, and performing potential entity relationship reasoning to generate potential relationship data;

[0123] In an embodiment of the present invention, the entity direct relationship data extracted in step S23 is subjected to relationship classification, and the relationships are divided into predefined relationship types, such as "treatment", "cause", "concurrent", etc. For example, the relationship "reason for taking" is classified as a "treatment" relationship. In addition, potential entity relationship reasoning is performed based on the medical knowledge graph or rules. For example, it is known that "hypertension" can "cause" "headache", and the patient has been identified as suffering from "hypertension". Even if "headache" is not explicitly mentioned in the text, it can be inferred that the patient has the potential symptom of "headache" and generate a potential relationship ("hypertension", "cause", "headache").

[0124] Step S25: performing weighted confidence evaluation on entity direct relationship data and potential relationship data, and performing medical term relationship mapping to generate medical term entity mapping relationship data;

[0125] In an embodiment of the present invention, the direct relationship data and potential relationship data of the entities obtained in steps S23 and S24 are subjected to confidence assessment. The confidence of direct relationships is usually high, while the confidence of potential relationships is relatively low. A weight can be assigned to each relationship according to the output probability of the relationship extraction model or a predefined rule to represent its confidence. For example, the weight of the direct relationship can be set to 0.9, while the weight of the potential relationship can be set to 0.5. Then, the entities and relationships are mapped to concepts in a standard medical terminology library (e.g., UMLS, SNOMEDCT). For example, "hypertension" is mapped to the concept of "primary hypertension", and "nifedipine" is mapped to its corresponding drug code.

[0126] Step S26: Perform triple entity relationship processing according to the medical term entity mapping relationship data to generate patient medical record relationship graph data.

[0127] In an embodiment of the present invention, the medical term entity mapping relationship data generated in step S25 is converted into a triple form (head entity, relationship, tail entity), for example ("primary hypertension", "treatment", "nifedipine"). Each triple represents an entity relationship in the medical record. All triples are combined to construct a medical record relationship map for the patient. The nodes in the map represent medical entities, the edges represent the relationships between entities, and the weights of the edges represent the confidence of the relationships. For example, "primary hypertension" and "nifedipine" can be used as nodes in the map, and they are connected by an edge labeled "treatment" and with a weight of 0.9.

[0128] Preferably, step S3 comprises the following steps:

[0129] Step S31: performing Bayesian network node processing based on the patient medical record relationship graph data to generate Bayesian network structure data;

[0130] Step S32: Deeply mine the patient's medication history through the time-series medical record medication event data to generate patient medication history feature data;

[0131] Step S33: performing a drug tolerance analysis on the patient according to the characteristic data of the patient's medication history, and generating the patient's drug tolerance characteristic data;

[0132] Step S34: using the patient's drug tolerance characteristic data to perform inter-node conditional probability processing on the Bayesian network structure data to generate Bayesian node conditional parameters;

[0133] Step S35: dynamically monitoring the patient's clinical physiological indicators according to the patient's drug tolerance characteristic data, and generating dynamic patient clinical physiological indicator data;

[0134] Step S36: Perform clinical physiological indicator risk probability reasoning on the dynamic patient clinical physiological indicator data through Bayesian node condition parameters to generate individual clinical medication risk data.

[0135] In an embodiment of the present invention, the entities in the patient medical record relationship graph data generated in step S2 are used as nodes of the Bayesian network. For example, diseases, symptoms, drugs, test results, etc. can all be used as nodes. The relationship between nodes determines the structure of the Bayesian network, that is, the connection mode and direction between nodes. For example, if disease A causes symptom B, then in the Bayesian network, node A points to node B. In addition, it is necessary to learn the structure of the network based on medical knowledge or data. For example, expert knowledge can be used to define the connection relationship between nodes, or a structural learning algorithm can be used to learn the network structure from data. The Bayesian network structure data finally generated contains node information and the connection relationship between nodes, for example: disease A->symptom B, drug C->disease A. Analyze the time series medical record medication event data generated in step S1 to extract the patient's medication history characteristics. For example, the number of times, dosage, duration, etc. of each drug used by the patient are counted. The patient's medication pattern can also be analyzed, such as whether there is long-term medication, combined medication, etc. For example, the patient used drug A 3 times, each dose was 100 mg, and the duration was 7 days; the patient used drug A and drug B at the same time. Convert this information into structured data, for example: {"Drug A":{"Number of uses":3, "Dosage":100, "Duration":7}, "Drug B":{"Number of uses":2, "Dosage":50, "Duration":5}, "Combination medication":["Drug A","Drug B"]}. Based on the patient's medication history feature data generated in step S32, combined with the patient's medical record information, analyze the patient's tolerance to different drugs. For example, the patient's tolerance to a drug can be judged based on whether the patient has adverse reactions after using a certain drug and how effective it is. For example, if a patient develops a rash after using drug A, it can be judged that the patient is intolerant to drug A. Quantify the patient's tolerance information for each drug, for example, use 0 for intolerance and 1 for tolerance. According to the patient's drug tolerance feature data generated in step S33, adjust the conditional probability between nodes in the Bayesian network. For example, if the patient is intolerant to drug A, the probability of drug A treating the disease can be reduced. The specific value of the conditional probability can be determined by expert knowledge or data learning methods. For example, according to clinical data statistics, the efficacy of drug A for tolerant people is 80%, and the efficacy for intolerant people is 30%. Then the conditional probability of drug A treating the disease in the Bayesian network can be set to 0.8 and 0.3 respectively. According to the patient's medication situation and tolerance characteristics, select the clinical physiological indicators that need to be monitored, such as blood pressure, heart rate, liver and kidney function, etc. Dynamically monitor these indicators and record the monitoring results to form time series data. For example, record the changes in blood pressure and heart rate of patients at different time points after taking drug A. Save these dynamic monitoring data as dynamic patient clinical physiological indicator data.For example, {"Time point 1": {"Blood pressure": 120 / 80, "Heart rate": 70}, "Time point 2": {"Blood pressure": 130 / 85, "Heart rate": 75}}. The dynamic patient clinical physiological indicator data generated in step S35 is input into the Bayesian network constructed in step S34, and the Bayesian network inference algorithm is used to calculate the probability of various clinical risks in patients. For example, the probability of patients experiencing adverse drug reactions, disease worsening and other risks can be calculated. For example, based on the patient's medication situation, tolerance characteristics, and dynamic changes in indicators such as blood pressure and heart rate, the probability of arrhythmia in the patient is calculated to be 0.1. These risk probabilities are used as individual clinical medication risk data, for example: {"Arrhythmia": 0.1, "Liver function damage": 0.05}.

[0136] Preferably, step S33 includes the following steps:

[0137] Step S331: classify the drugs according to the patient's medication history characteristic data to obtain the patient's classified medication data;

[0138] Step S332: analyzing the drug dosage change of the patient's classified medication data to obtain the patient's medication dosage change data;

[0139] Step S333: Evaluate the efficacy index value of the patient's medication history characteristic data through the patient's classified medication data to generate the patient's medication efficacy index value;

[0140] Step S334: using the patient's medication dosage change data to process the patient's medication efficacy index value into a therapeutic effect response curve, and performing a drug sensitivity analysis to obtain drug sensitivity assessment data;

[0141] Step S335: Calculate the dose adjustment factor required for the target therapeutic effect according to the drug sensitivity assessment data, and perform drug tolerance numerical label processing to obtain numerical tolerance data;

[0142] Step S336: Integrate the numerical tolerance data, the patient's medication dosage change data, and the patient's medication efficacy index value to generate the patient's medication tolerance characteristic data.

[0143] In an embodiment of the present invention, the patient's medication history characteristic data is obtained from step S32, which contains all the drug information used by the patient. The drugs are classified according to the pharmacological effects of the drugs, the therapeutic field or the ATC classification system. For example, the drugs can be divided into categories such as antibiotics, antihypertensive drugs, and hypoglycemic drugs. Assuming that the patient's medication history characteristic data contains drugs A, B, and C, where A and B are antibiotics and C is an antihypertensive drug, the patient's classified medication data is {"antibiotics": [A, B], "antihypertensive drugs": [C]}. For each type of drug, the change trend of the patient's medication dosage is analyzed. For example, the difference between each medication dosage and the previous medication dosage can be calculated, or the average, maximum, minimum and other statistical indicators of the medication dosage over a period of time can be calculated. Assuming that the patient's dosage record of antibiotic A is [100mg, 150mg, 200mg], the dosage change can be calculated as [+50mg, +50mg], or the average dosage can be calculated as 150mg. These dosage change information are recorded as patient dosage change data, for example: {"Antibiotic A":{"Dose change":[+50mg,+50mg],"Average dose":150mg},"Antibiotic B":{...}}. According to the treatment purpose corresponding to each type of drug in the patient classification medication data, the corresponding efficacy index is selected for evaluation. For example, for antibiotics, the changes in infection indicators (such as white blood cell count and body temperature) can be evaluated; for antihypertensive drugs, the changes in blood pressure can be evaluated. Assuming that the white blood cell count of the patient drops from 15x10^9 / L to 8x10^9 / L during the use of antibiotic A, it can be considered that drug A has a significant improvement in the patient's infection indicators. These efficacy index values ​​are recorded to form patient drug efficacy index values, for example: {"Antibiotic A":{"White blood cell count":[15,12,8]}, "Antihypertensive drug C":{"Blood pressure":[160 / 100,150 / 90,140 / 80]}}. Combine the patient's medication dosage change data and the patient's drug efficacy index value to draw an efficacy response curve. For example, the dose of antibiotic A can be used as the horizontal axis and the white blood cell count as the vertical axis to draw a scatter plot and fit the curve. By observing the slope, inflection point and other characteristics of the curve, the patient's sensitivity to the drug can be judged. For example, if the slope of the curve is large, it means that the patient is more sensitive to the drug; if the slope of the curve is small, or even a plateau occurs, it means that the patient is less sensitive to the drug. Quantify the results of the drug sensitivity analysis, for example, use a numerical value between 0 and 1 to represent the degree of sensitivity, and form drug sensitivity assessment data, for example: {"antibiotic A": 0.8, "antihypertensive drug C": 0.6}. Based on the drug sensitivity assessment data, calculate the dose adjustment factor required to achieve the target efficacy. For example, if the patient is less sensitive to drug A, the dose needs to be increased to achieve the target efficacy.Assuming that the patient's sensitivity to drug A is 0.6 and the target efficacy requires a sensitivity of 0.8, the dose adjustment factor can be calculated as 0.8 / 0.6=1.33. Convert the drug sensitivity assessment data into a numerical tolerance label. For example, the sensitivity value can be directly used as the tolerance value, or converted according to a predefined rule. For example, a drug with a sensitivity value greater than 0.7 can be marked as tolerant (value 1), and a drug with a sensitivity value less than 0.7 can be marked as intolerant (value 0). The data generated in steps S332, S333, and S335 are integrated together to form the final patient drug tolerance feature data. The data contains the patient's tolerance numerical label for each drug, the dosage change information, and the efficacy index value, for example: {"Antibiotic A":{"Tolerance":1, "Dose Change":[+50mg,+50mg], "Average Dose":150mg, "White Blood Cell Count":[15,12,8]},"Antihypertensive Drug C":{...}}.

[0144] Preferably, step S35 includes the following steps:

[0145] Step S351: Processing a key monitoring organ list according to the patient's drug tolerance characteristic data to generate key monitoring organ data;

[0146] Step S352: configuring monitoring indicators for key monitoring organ data through a preset list of monitoring organ physiological indicators to generate monitoring indicator configuration data;

[0147] Step S353: monitoring the patient's organ physiological indexes using clinical monitoring equipment based on the monitoring index configuration data to generate clinical organ monitoring physiological index data;

[0148] Step S354: performing organ function status classification on the clinical organ monitoring physiological index data to obtain organ function status classification data;

[0149] Step S355: Based on the organ function status classification data, the physiological index change rate of the clinical organ monitoring physiological index data is processed to obtain dynamic patient clinical physiological index data.

[0150] In an embodiment of the present invention, the patient's drug tolerance characteristic data is obtained from step S33, which includes the patient's tolerance information to different drugs. According to the pharmacological effects and potential adverse reactions of the drug, the organs that need to be monitored are determined. For example, if the patient is taking a drug that is potentially toxic to the liver and has low tolerance, the liver needs to be listed as a key monitoring organ. For another example, if the patient has low tolerance to a certain antihypertensive drug, it is necessary to focus on monitoring heart and kidney function. Assuming that the patient is taking drug A that is potentially toxic to the liver and kidneys, the key monitoring organ data is ["liver", "kidney"]. According to the key monitoring organs determined in step S351, the corresponding physiological indicators are selected from the preset monitoring organ physiological indicator list for monitoring. The preset monitoring organ physiological indicator list contains physiological indicators corresponding to different organs, for example, liver indicators include alanine aminotransferase (ALT), aspartate aminotransferase (AST), total bilirubin, etc.; kidney indicators include creatinine, urea nitrogen, uric acid, etc. Assuming that the key monitored organs are the liver and kidneys, the monitoring index configuration data is {"liver":["ALT","AST","total bilirubin"], "kidney":["creatinine","urea nitrogen"]}. Use clinical monitoring equipment, such as a biochemical analyzer, an electrocardiograph, a sphygmomanometer, etc., to monitor the physiological indicators configured in step S352 in real time. For example, use a biochemical analyzer to detect the patient's ALT, AST, total bilirubin, creatinine, urea nitrogen and other indicators. Record the monitored data and associate it with the corresponding timestamp. For example, at 10:00, the ALT value is 40U / L and the AST value is 35U / L. Save these data as clinical organ monitoring physiological indicator data, for example: {"liver":{"10:00":{"ALT":40,"AST":35,"total bilirubin":15},"12:00":{"ALT":45,"AST":40,"total bilirubin":16}},"kidney":{...}}. Use clinical monitoring equipment, such as a biochemical analyzer, an electrocardiograph, a sphygmomanometer, etc., to monitor the physiological indicators configured in step S352 in real time. For example, use a biochemical analyzer to detect the patient's ALT, AST, total bilirubin, creatinine, urea nitrogen and other indicators. Record the monitored data and associate it with the corresponding timestamp. For example, at 10:00, the ALT value is 40U / L and the AST value is 35U / L. These data are saved as clinical organ monitoring physiological indicator data, for example: {"liver":{"10:00":{"ALT":40,"AST":35,"total bilirubin":15},"12:00":{"ALT":45,"AST":40,"total bilirubin":16}},"kidney":{...}}. According to clinical guidelines or expert experience, the clinical organ monitoring physiological indicator data obtained in step S353 are graded for organ function status.For example, liver function can be divided into four levels: normal, mildly abnormal, moderately abnormal, and severely abnormal according to the values ​​of ALT and AST. Assuming that the patient's ALT value is 40U / L and the AST value is 35U / L, according to the predefined rules, its liver function status can be classified as mildly abnormal. The functional status classification results of each key monitored organ are recorded to form organ function status classification data, for example: {"liver":{"10:00":"mildly abnormal","12:00":"mildly abnormal"}, "kidney":{"10:00":"normal","12:00":"normal"}}. Calculate the change rate of the clinical organ monitoring physiological indicator data obtained in step S353, and combine the change rate with the organ function status classification data obtained in step S354 to generate dynamic patient clinical physiological indicator data. For example, the percentage change of the physiological indicator at each time point relative to the baseline value can be calculated, or the change amount of the physiological indicator between two adjacent time points can be calculated. Assuming that the patient's ALT value at 10:00 is 40 U / L and the ALT value at 12:00 is 45 U / L, the change rate of ALT is (45-40) / 40 = 0.125. The physiological index value, change rate and organ function status classification are integrated together to form the final dynamic patient clinical physiological index data.

[0151] As an example of the present invention, refer to Figure 3 Show, for Figure 1 The detailed implementation step flow chart of step S4 in the embodiment includes:

[0152] Step S41: Evaluate the metabolic capacity of the patient according to the individual clinical medication risk data to obtain metabolic capacity score data of the patient;

[0153] In an embodiment of the present invention, individual clinical medication risk data is obtained from step S3, which includes the probability of various clinical risks of the patient, such as liver function impairment, renal function impairment, arrhythmia, etc. Based on these risk probabilities, combined with a pre-set risk-metabolic capacity mapping rule or model, the patient's metabolic capacity is evaluated, and the patient's metabolic capacity score data is generated. For example, a model based on Logistic regression can be constructed, using various risk probabilities as input features, and outputting the patient's metabolic capacity score (e.g., 0-10 points, the higher the score, the better the metabolic capacity). Assuming that a patient's liver function impairment risk probability is 0.2, the renal function impairment risk probability is 0.1, and other risk probabilities are all low, the model predicts that the patient's metabolic capacity score is 8 points. Then the patient's metabolic capacity score data is 8.

[0154] Step S42: Calculating the initial adjusted drug clearance rate based on the patient's metabolic capacity score data to generate initial drug adjusted drug clearance rate data;

[0155] In an embodiment of the present invention, the initial adjusted clearance rate of the drug is calculated based on the patient's metabolic capacity score data obtained in step S41. The calculation can be performed according to a predefined score-clearance mapping rule or model. For example, a linear model can be set to multiply the metabolic capacity score by the standard clearance rate to obtain the initial adjusted clearance rate. Assuming that the standard clearance rate of a certain drug is 50 mL / min and the patient's metabolic capacity score is 8 points (out of 10 points), the initial adjusted clearance rate is 50 mL / min×(8 / 10)=40 mL / min. The calculated initial adjusted clearance rate is saved as the initial drug adjusted clearance rate data.

[0156] Step S43: Acquire the patient's renal function index data and the patient's liver function index data;

[0157] In an embodiment of the present invention, renal function index data and liver function index data are obtained from the patient's electronic medical record or test results. The renal function index data may include serum creatinine, creatinine clearance, urea nitrogen, etc.; the liver function index data may include alanine aminotransferase (ALT), aspartate aminotransferase (AST), total bilirubin, albumin, etc. For example, the patient's serum creatinine value is 1.2 mg / dL, the ALT value is 30 U / L, and the AST value is 25 U / L.

[0158] Step S44: Based on the patient's renal function index data and the patient's liver function index data, the initial drug-adjusted clearance rate data is corrected for liver and kidney baseline values ​​to generate corrected drug metabolism clearance baseline data.

[0159] In an embodiment of the present invention, the initial drug-adjusted clearance calculated in step S42 is corrected according to the patient's renal function index data and liver function index data obtained in step S43. The correction method can be based on a pharmacokinetic model or a clinical experience formula. For example, for drugs that are mainly excreted by the kidneys, the Cockcroft-Gault formula can be used to calculate the creatinine clearance based on the patient's age, weight, gender and serum creatinine value, and then the initial drug-adjusted clearance is adjusted according to the ratio of the creatinine clearance to the standard clearance. Assuming that the patient's creatinine clearance calculated according to the Cockcroft-Gault formula is 60 mL / min and the standard clearance of the drug is 75 mL / min, the corrected drug clearance is 40 mL / min×(60 mL / min / 75 mL / min)=32 mL / min. For drugs that are mainly metabolized by the liver, the initial drug-adjusted clearance can be corrected according to liver function indicators, such as the Child-Pugh score. Assuming that the patient's Child-Pugh score is 5 (mild liver function impairment), according to the pre-set rule, the initial drug adjustment clearance is reduced by 20%, and the corrected drug clearance is 40mL / min×(1-0.2)=32mL / min. The final corrected clearance is saved as the corrected drug metabolism clearance baseline data.

[0160] Preferably, step S5 comprises the following steps:

[0161] Step S51: querying the clinical disease-effect drugs of the patient through a preset drug database to generate clinical physiological effect drug data;

[0162] Step S52: adjusting the individual characteristic drug dosage of the clinical physiological effect drug data by correcting the drug metabolism clearance baseline data to obtain personalized dosage range data;

[0163] Step S53: Based on the personalized dose range data, the patient's drug dose response is simulated through the clinical physiological drug data, and the response index is predicted to obtain the predicted drug concentration data, the predicted efficacy index value and the adverse reaction probability data at different doses;

[0164] Step S54: performing drug dose curve axis processing on the predicted drug concentration data, predicted efficacy index value and adverse reaction probability data through the personalized dose range data to generate dose response curve data;

[0165] Step S55: Perform interactive control processing according to the dose response curve data, and perform dynamic visualization of the curve to generate visualized clinical medication response data.

[0166] In an embodiment of the present invention, based on the patient's diagnosis information, such as the disease information extracted from the patient's medical history relationship map generated in step S2, a preset drug database (such as DrugBank, PharmGKB) is queried for drugs that can be used to treat the disease. For example, if the patient is diagnosed with hypertension, some commonly used antihypertensive drugs, such as nifedipine, captopril, etc., can be queried. The clinical physiological effect drug data includes information such as the name, pharmacological action, usage and dosage, adverse reactions, and pharmacokinetic parameters of the drug. For example, the clinical physiological effect drug data of nifedipine includes: name: nifedipine; pharmacological action: calcium channel blocker; usage and dosage: oral, 10 mg, three times a day; adverse reactions: headache, edema; pharmacokinetic parameters: half-life is 2-5 hours, mainly metabolized by the liver. According to the corrected drug metabolism clearance baseline data generated in step S4, the recommended dose in the clinical physiological effect drug data obtained in step S51 is adjusted. The adjustment method can be based on a pharmacokinetic model or a clinical experience formula. For example, the maintenance dose of the drug can be adjusted according to the patient's clearance rate, or the initial dose of the drug can be adjusted according to the patient's liver and kidney function. Assuming that the recommended dose of a drug is 10 mg, three times a day, and the patient's modified clearance is 80% of the standard clearance, the dose can be adjusted to 10 mg × 0.8 = 8 mg, three times a day. In order to provide more flexible dosage options, a personalized dose range can be generated instead of a single dose value. For example, depending on the patient's specific situation, the dose range can be set to 6 mg-10 mg, three times a day. Using a pharmacokinetic / pharmacodynamic (PK / PD) model, combined with clinical physiological drug data and personalized dose range data, simulate the patient's medication response at different doses. For example, the changes in drug concentration in the body, changes in efficacy indicators, and the probability of adverse reactions at different doses can be simulated. Assuming that the personalized dose range is 6 mg-10 mg, the medication response at three doses of 6 mg, 8 mg, and 10 mg can be simulated. The simulation results can obtain predicted drug concentration data (for example, drug plasma concentration at different time points), predicted efficacy index values ​​(for example, blood pressure, heart rate, etc.) and adverse reaction probability data (for example, the probability of occurrence of headache and edema) at different doses. The simulation results generated in step S53 are presented in the form of a dose response curve. For example, the drug dose can be used as the abscissa, and the drug concentration, efficacy index or adverse reaction probability can be used as the ordinate to draw a curve. Multiple curves can be drawn to represent drug concentrations at different time points, different efficacy indexes or different adverse reactions. For example, three curves can be drawn to represent the changes in drug concentration over time at three doses of 6 mg, 8 mg and 10 mg. These curve data are saved as dose response curve data. The dose response curve data generated in step S54 is visualized.You can use chart libraries (such as matplotlib, plotly) to draw graphs and add interactive controls such as sliders and drop-down menus so that users can dynamically adjust the dose and view the corresponding simulation results in real time. For example, users can adjust the drug dose through the slider, and the graph will be updated in real time to show the changes in drug concentration, efficacy indicators and adverse reaction probability at different doses. Integrate the visualized dose-response curve with relevant clinical information (such as basic patient information, disease diagnosis, and drug information) to generate visualized clinical medication response data, providing clinicians with intuitive medication guidance.

[0167] The present invention also provides a clinical internal medicine visualization system, which executes the above-mentioned clinical internal medicine visualization method, and the clinical internal medicine visualization system comprises:

[0168] The medical record data preprocessing module is used to obtain the basic information of clinical patients; segment the medical record terminology text according to the basic information of clinical patients to generate medical record text terminology segmentation data; analyze the medical treatment and medication events of the medical record text terminology segmentation data to generate time series medical record medication event data;

[0169] The entity relationship graph module is used to perform medical entity recognition on the medical record text term segmentation data to obtain medical entity data; perform medical term relationship mapping based on the medical entity data to generate medical term entity mapping relationship data; perform triple entity relationship processing based on the medical term entity mapping relationship data to generate patient medical record relationship graph data;

[0170] The tolerance index risk module is used to process Bayesian network nodes based on the patient medical record relationship map data to generate Bayesian network structure data; perform patient drug tolerance analysis based on the time-series medical record medication event data to generate patient drug tolerance characteristic data; dynamically monitor the patient's clinical physiological indicators based on the patient's drug tolerance characteristic data to generate dynamic patient clinical physiological indicator data; perform clinical physiological indicator risk probability reasoning on the Bayesian network structure data based on the dynamic patient clinical physiological indicator data using the patient's drug tolerance characteristic data to generate individual clinical medication risk data;

[0171] The metabolic capacity assessment module is used to assess the metabolic capacity of patients based on individual clinical medication risk data and generate modified drug metabolic clearance baseline data;

[0172] The dose adjustment visualization module is used to obtain clinical physiological drug data; by correcting the drug metabolism clearance baseline data, the clinical physiological drug data is adjusted for individual characteristic drug doses, and the drug dose curve is visualized and dynamically displayed to generate visualized clinical medication response data.

[0173] The present application is to obtain basic information of clinical patients, and combine the steps of medical record term text segmentation, medical treatment and medication event analysis, so as to fully and deeply explore and understand the patient's condition and medical history. Traditional clinical medication decisions mainly rely on the doctor's experience and memory, and the method of the present application can structure and digitize the patient's medical record information, so that the doctor can quickly and comprehensively grasp the patient's health status and medication history. For example, by segmenting and entity recognition of the medical record text, the system can accurately extract key information such as the patient's disease diagnosis, symptom manifestations, and previous medication. This helps doctors avoid medication decision errors caused by information omissions or memory bias, and improves the scientificity and rationality of the medication plan. According to the individual characteristics and metabolic capacity of the patient, the drug dosage is accurately adjusted, and the drug concentration, efficacy index and adverse reaction probability at different doses are intuitively displayed through the dynamic display of the drug dosage curve.

[0174] Therefore, the embodiments should be regarded as illustrative and non-restrictive from all points, and the scope of the present invention is limited by the appended claims rather than the above description, and it is therefore intended that all changes falling within the meaning and range of equivalent elements of the application documents are included in the present invention.

[0175] The above description is only a specific embodiment of the present invention, so that those skilled in the art can understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but should conform to the widest scope consistent with the principles and novel features invented herein.

Claims

1. A visualization method for clinical internal medicine, characterized in that: The following steps are involved: Step S1: Obtain basic information of clinical patients; segment medical record terminology text according to the basic information of clinical patients to generate medical record text terminology segmentation data; perform medical treatment and medication event analysis on the medical record text terminology segmentation data to generate time series medical record medication event data; Step S2: performing medical entity recognition on the medical record text term segmentation data to obtain medical entity data; Perform medical terminology relationship mapping based on medical entity data to generate medical terminology entity mapping relationship data; Perform triple entity relationship processing based on medical term entity mapping relationship data to generate patient medical record relationship graph data; Step S3: Perform Bayesian network node processing based on the patient medical record relationship graph data to generate Bayesian network structure data; Conduct patient drug tolerance analysis based on time-series medical record medication event data to generate patient drug tolerance feature data; Dynamically monitor the patient's clinical physiological indicators based on the patient's drug tolerance characteristic data and generate dynamic patient clinical physiological indicator data; Based on the dynamic patient clinical physiological index data, the patient's drug tolerance characteristic data is used to perform clinical physiological index risk probability inference on the Bayesian network structure data to generate individual clinical medication risk data; Step S4: Evaluate the patient's metabolic capacity based on individual clinical medication risk data and generate revised drug metabolic clearance baseline data; Step S5: Acquiring clinical physiological effect drug data; By correcting the drug metabolism and clearance baseline data, the clinical physiological effect drug data is adjusted for individual characteristic drug dosage, and the drug dosage curve is visualized and dynamically displayed to generate visualized clinical medication response data.

2. The clinical internal medicine visualization method according to claim 1, characterized in that: Step S1 includes the following steps: Step S11: Obtaining basic clinical patient information; Step S12: extracting historical medical records based on the basic information of clinical patients to generate historical medical record data of clinical patients; Step S13: performing historical data cleaning processing according to the historical medical record data of clinical patients, and performing terminology text segmentation on the electronic medical record standard data through a preset medical term dictionary to generate medical record text term segmentation data; Step S14: Analyze the medical record text term segmentation data for medical treatment and medication events, and perform time axis alignment processing to generate time-series medical record medication event data.

3. The clinical internal medicine visualization method according to claim 2, characterized in that: Step S13 includes the following steps: Step S131: Perform text encoding format detection on clinical patient historical medical record data to obtain medical record encoding data; Step S132: decoding the historical medical record data of clinical patients through the medical record coding data to generate decoded medical record text data; Step S133: using a preset regular expression to match special symbols in the decoded medical record text data, and removing non-medical text to obtain simplified historical medical record text data; Step S134: matching the separator positions of the simplified historical medical record text data with the preset medical paragraph separators, and performing separator position analysis to generate separator position data; Step S135: segmenting the simplified historical medical record text data into text segments according to the separator position data to obtain historical medical record segmentation data; Step S136: Use a preset medical term dictionary to perform paragraph type keyword matching on the historical medical record paragraph segmentation data to obtain paragraph type label data; Step S137: Utilize the paragraph type label data to merge the same type of paragraphs in the simplified historical medical record text data to generate medical record text term segmentation data.

4. The clinical internal medicine visualization method according to claim 1, characterized in that: Step S2 includes the following steps: Step S21: performing medical entity recognition on the medical record text term segmentation data to obtain medical entity data; Step S22: performing entity disambiguation on the medical entity data, and using a pre-trained medical relationship extraction model to extract the relationship between entities to obtain dependency tree data; Step S23: performing entity direct relationship analysis according to the dependency tree data to obtain entity direct relationship data; Step S24: classifying the relationships according to the entity direct relationship data, and performing potential entity relationship reasoning to generate potential relationship data; Step S25: performing weighted confidence evaluation on entity direct relationship data and potential relationship data, and performing medical term relationship mapping to generate medical term entity mapping relationship data; Step S26: Perform triple entity relationship processing according to the medical term entity mapping relationship data to generate patient medical record relationship graph data.

5. The clinical internal medicine visualization method according to claim 1, characterized in that: Step S3 includes the following steps: Step S31: performing Bayesian network node processing based on the patient medical record relationship graph data to generate Bayesian network structure data; Step S32: Deeply mine the patient's medication history through the time-series medical record medication event data to generate patient medication history feature data; Step S33: performing a drug tolerance analysis on the patient according to the characteristic data of the patient's medication history, and generating the patient's drug tolerance characteristic data; Step S34: using the patient's drug tolerance characteristic data to perform inter-node conditional probability processing on the Bayesian network structure data to generate Bayesian node conditional parameters; Step S35: dynamically monitoring the patient's clinical physiological indicators according to the patient's drug tolerance characteristic data, and generating dynamic patient clinical physiological indicator data; Step S36: Perform clinical physiological indicator risk probability reasoning on the dynamic patient clinical physiological indicator data through Bayesian node condition parameters to generate individual clinical medication risk data.

6. The clinical internal medicine visualization method according to claim 5, characterized in that: Step S33 includes the following steps: Step S331: classify the drugs according to the patient's medication history characteristic data to obtain the patient's classified medication data; Step S332: analyzing the drug dosage change of the patient's classified medication data to obtain the patient's medication dosage change data; Step S333: Evaluate the efficacy index value of the patient's medication history characteristic data through the patient's classified medication data to generate the patient's medication efficacy index value; Step S334: using the patient's medication dosage change data to process the patient's medication efficacy index value into a therapeutic effect response curve, and performing a drug sensitivity analysis to obtain drug sensitivity assessment data; Step S335: Calculate the dose adjustment factor required for the target therapeutic effect according to the drug sensitivity assessment data, and perform drug tolerance numerical label processing to obtain numerical tolerance data; Step S336: Integrate the numerical tolerance data, the patient's medication dosage change data, and the patient's medication efficacy index value to generate the patient's medication tolerance characteristic data.

7. The clinical internal medicine visualization method according to claim 5, characterized in that: Step S35 includes the following steps: Step S351: Processing a key monitoring organ list according to the patient's drug tolerance characteristic data to generate key monitoring organ data; Step S352: configuring monitoring indicators for key monitoring organ data through a preset list of monitoring organ physiological indicators to generate monitoring indicator configuration data; Step S353: monitoring the patient's organ physiological indexes using clinical monitoring equipment based on the monitoring index configuration data to generate clinical organ monitoring physiological index data; Step S354: performing organ function status classification on the clinical organ monitoring physiological index data to obtain organ function status classification data; Step S355: Based on the organ function status classification data, the physiological index change rate of the clinical organ monitoring physiological index data is processed to obtain dynamic patient clinical physiological index data.

8. The clinical internal medicine visualization method according to claim 1, characterized in that: Step S4 includes the following steps: Step S41: Evaluate the metabolic capacity of the patient according to the individual clinical medication risk data to obtain metabolic capacity score data of the patient; Step S42: Calculating the initial adjusted drug clearance rate based on the patient's metabolic capacity score data to generate initial drug adjusted drug clearance rate data; Step S43: Acquire the patient's renal function index data and the patient's liver function index data; Step S44: Based on the patient's renal function index data and the patient's liver function index data, the initial drug-adjusted clearance rate data is corrected for liver and kidney baseline values ​​to generate corrected drug metabolism clearance baseline data.

9. The clinical internal medicine visualization method according to claim 1, characterized in that: Step S5 includes the following steps: Step S51: querying the clinical disease-effect drugs of the patient through a preset drug database to generate clinical physiological effect drug data; Step S52: adjusting the individual characteristic drug dosage of the clinical physiological effect drug data by correcting the drug metabolism clearance baseline data to obtain personalized dosage range data; Step S53: Based on the personalized dose range data, the patient's drug dose response is simulated through the clinical physiological drug data, and the response index is predicted to obtain the predicted drug concentration data, the predicted efficacy index value and the adverse reaction probability data at different doses; Step S54: performing drug dose curve axis processing on the predicted drug concentration data, predicted efficacy index value and adverse reaction probability data through the personalized dose range data to generate dose response curve data; Step S55: Perform interactive control processing according to the dose response curve data, and perform dynamic visualization of the curve to generate visualized clinical medication response data.

10. A visualization system for clinical internal medicine, characterized in that: For executing the clinical internal medicine visualization method according to claim 1, the clinical internal medicine visualization system comprises: The medical record data preprocessing module is used to obtain the basic information of clinical patients; segment the medical record terminology text according to the basic information of clinical patients to generate medical record text terminology segmentation data; analyze the medical treatment and medication events of the medical record text terminology segmentation data to generate time-series medical record medication event data; The entity relationship graph module is used to perform medical entity recognition on the medical record text term segmentation data to obtain medical entity data; perform medical term relationship mapping based on the medical entity data to generate medical term entity mapping relationship data; perform triple entity relationship processing based on the medical term entity mapping relationship data to generate patient medical record relationship graph data; The tolerance index risk module is used to process Bayesian network nodes based on the patient medical record relationship graph data to generate Bayesian network structure data; analyze the patient's drug tolerance based on the time-series medical record medication event data to generate the patient's drug tolerance characteristic data; dynamically monitor the patient's clinical physiological indicators based on the patient's drug tolerance characteristic data to generate dynamic patient clinical physiological indicator data; perform clinical physiological indicator risk probability reasoning on the Bayesian network structure data based on the dynamic patient clinical physiological indicator data using the patient's drug tolerance characteristic data to generate individual clinical medication risk data; The metabolic capacity assessment module is used to assess the metabolic capacity of patients based on individual clinical medication risk data and generate modified drug metabolic clearance baseline data; The dose adjustment visualization module is used to obtain clinical physiological drug data; by correcting the drug metabolism clearance baseline data, the clinical physiological drug data is adjusted for individual characteristic drug doses, and the drug dose curve is visualized and dynamically displayed to generate visualized clinical medication response data.

Images