Medical term analysis method based on knowledge graph technology

Abstract

The invention discloses a medical term analysis method based on a knowledge graph technology, and the method comprises the following steps: obtaining medical standard literatures of different time versions, and constructing a medical term knowledge graph sequence containing time dimension information; based on the medical term knowledge graph sequence, calculating semantic representation changes of the target medical term at different time points to obtain a semantic evolution path of the target medical term; acquiring clinical disease course data related to the target medical term; based on a statistical significance test method, calculating the support degree of the clinical disease course data to emerging semantics in the semantic evolution path, and generating a quantitative verification result; and dynamically updating and optimizing the medical term knowledge graph sequence according to the quantitative verification result. According to the method, prospective active perception and accurate verification of emerging semantics are realized, and the maintenance mode of the knowledge graph can be converted into active optimization from post passive updating, so that potential medical risks caused by indefinite term evolution are effectively reduced.

Description

Technical Field

[0001] This invention relates to the field of medical information processing technology, and in particular to a method for analyzing medical terminology based on knowledge graph technology. Background Technology

[0002] Currently, medical terminology is widely used in clinical practice, research literature, and standard guidelines. However, with the continuous updating of medical knowledge and the ongoing development of clinical practice, the semantic meaning of medical terms often evolves over time. For example, the same term may undergo semantic strengthening, weakening, ambiguity, or even radical changes at different periods. If this semantic drift is not captured and verified in a timely manner, it may lead to lag or distortion in the conceptual associations in the medical knowledge graph, thereby affecting the accuracy of clinical decision support and even bringing potential medical risks.

[0003] In existing technologies, the maintenance of medical knowledge graphs largely relies on manual review and post-hoc revision, lacking the ability to systematically quantify and predict the semantic evolution of terms. On the one hand, traditional methods struggle to reveal the semantic change trends of terms over time; on the other hand, even if emerging semantics are detected, there is a lack of mechanisms for quantitative verification using clinical data, making it difficult to ensure the scientific validity and reliability of knowledge graph updates. Summary of the Invention

[0004] The purpose of this section is to outline some aspects of embodiments of the present invention and to briefly describe some preferred embodiments. Simplifications or omissions may be made in this section, as well as in the abstract and title of this application, to avoid obscuring the purpose of these documents; however, such simplifications or omissions should not be construed as limiting the scope of the invention.

[0005] To address the aforementioned technical problems, this invention provides the following technical solution: a medical terminology analysis method based on knowledge graph technology, comprising the following steps: Obtain medical standard documents from different time periods and construct a sequence of medical terminology knowledge graphs containing time-related information. Based on the medical terminology knowledge graph sequence, the semantic representation changes of the target medical term at different points in time are calculated to obtain the semantic evolution path of the target medical term. Acquire clinical course data related to the target medical term; calculate the support of the clinical course data for emerging semantics in the semantic evolution path based on statistical significance testing methods, and generate quantitative verification results; Based on the quantitative verification results, the medical terminology knowledge graph sequence is dynamically updated and optimized.

[0006] As a preferred embodiment of the medical terminology analysis method based on knowledge graph technology described in this invention, the construction of a medical terminology knowledge graph sequence containing time dimension information includes the following steps: Knowledge is extracted from medical literature at each time period to form multiple static knowledge graphs; Each static knowledge graph is assigned a timestamp and organized in chronological order to form the knowledge graph sequence.

[0007] As a preferred embodiment of the medical terminology analysis method based on knowledge graph technology described in this invention, the method involves: calculating the semantic representation changes of the target medical term at different points in time based on the medical terminology knowledge graph sequence to obtain the semantic evolution path of the target medical term, including the following steps: For each of the static knowledge graphs in the medical terminology knowledge graph sequence, a node sequence conforming to the knowledge distribution at the corresponding time point is generated by a random walk algorithm, so as to construct an independent time-series text corpus for each time point; A word embedding model is used, which is trained independently on an independent temporal text corpus corresponding to each time point to obtain the vector representation of all terms at each time point; By employing vector space alignment technology, term vectors at different time points are mapped to a unified vector space, making term vectors at different time points comparable. For a target medical term, its spatially aligned vector representation at consecutive time points is extracted, and the degree of drift of the term vector between two time points is quantified by calculating the cosine distance or Euclidean distance. By connecting the semantic drift of the target term at all consecutive time points in chronological order, a curve showing the change of its semantic similarity over time is plotted. By analyzing the trend of this path, the semantic evolution pattern of the term is determined, thus constructing the semantic evolution path of the target medical term.

[0008] As a preferred embodiment of the medical terminology analysis method based on knowledge graph technology described in this invention, after obtaining the semantic evolution path of the target medical term, a prediction and early warning step is further included: Based on the degree of drift of the term vector, a time series prediction algorithm is used to predict the semantic vector representation of the target medical term at future time points; Based on the prediction results, a semantic evolution trend warning for the term is generated, which includes semantic strengthening, semantic weakening, semantic divergence, or semantic subversion.

[0009] As a preferred embodiment of the medical terminology analysis method based on knowledge graph technology described in this invention, the step of generating the semantic evolution trend warning is as follows: Based on the predicted future semantic vector, calculate the cosine or Euclidean distance between it and the latest current vector; If the distance is significantly greater than the historical average drift, a semantic subversion warning is generated, indicating that the meaning of the term may undergo a major change, and suggesting that the new meaning represented by the direction of this change is a potential emerging semantic that needs to be verified. If the distance between the predicted vector and the vector cluster representing other semantic concepts decreases significantly, a semantic disambiguation warning is generated, indicating that the term may derive a new semantic branch and suggesting that the new branch is a potential emerging semantic to be verified. If the distance change shows a stable positive or negative trend, a corresponding warning of semantic strengthening or weakening will be generated, indicating that the semantic association between the term and a specific diagnostic concept or entity is strengthening or weakening, and suggesting that the strengthened or weakened association is a potential emerging semantic to be verified.

[0010] As a preferred embodiment of the medical terminology analysis method based on knowledge graph technology described in this invention, the method includes: calculating the support of clinical course data for emerging semantics in the semantic evolution path based on statistical significance testing, including: Clinical course data of treatment plans that adopted the new semantics were extracted as the experimental group, and clinical course data of traditional treatment plans were extracted as the control group; wherein, the clinical course data includes, but is not limited to, diagnostic information, medication records, surgical procedures, laboratory test results and efficacy evaluation records; The differences between the experimental group and the control group in terms of preset clinical efficacy indicators are quantitatively calculated; the preset clinical efficacy indicators include at least one of survival rate, adverse reaction rate, and symptom relief period. A statistical hypothesis testing model is constructed. The model takes the efficacy index data of the experimental group and the control group as input, calculates and obtains the statistical measure representing the significance of the difference, and outputs the corresponding p value.

[0011] As a preferred embodiment of the medical terminology analysis method based on knowledge graph technology described in this invention, the construction of the statistical hypothesis testing model includes: A multivariate statistical significance test model was constructed to control for the interference of patient baseline characteristics differences on efficacy indicators; the multivariate statistical significance test model was a logistic regression model or a Cox proportional hazards model. The dependent variable of the model is a preset clinical efficacy indicator; The core independent variable of the model is the treatment regimen grouping; Patient age, gender, and disease stage are included as covariates in the model. By fitting the model, the regression coefficients, Wald statistic, and p-values ​​corresponding to the core independent variables are calculated. The p-value is the significance measure of the effect of treatment grouping on efficacy after excluding the interference of covariates.

[0012] As a preferred embodiment of the medical terminology analysis method based on knowledge graph technology described in this invention, generating quantitative verification results includes: The preset significance level threshold is ; Compare the calculated p-value with the threshold Compare; If p value < Then, a quantitative verification result of successful clinical validation is generated; If p value ≥ This will generate a quantitative validation result indicating that the clinical validation was unsuccessful.

[0013] As a preferred embodiment of the medical terminology analysis method based on knowledge graph technology described in this invention, generating quantitative verification results includes: dynamically updating and optimizing the medical terminology knowledge graph sequence based on the quantitative verification results, including: If the quantitative verification result indicates successful clinical validation, then the first optimization strategy is automatically executed, which includes: Based on the p-value of the statistical significance test, the edge weight between the target medical term and the treatment plan node corresponding to the emerging semantic is increased according to the preset first update rule. Add clinical evidence support annotations to this emerging semantic node in the knowledge graph and record the verification timestamp for traceability and subsequent version management; If the quantitative verification result indicates that the clinical verification was unsuccessful, a second optimization strategy will be automatically executed. The second optimization strategy includes: Based on the p-value of the statistical significance test, the association confidence between the target medical term and the treatment plan node corresponding to the emerging semantics is reduced according to the preset second update rule. If the association confidence level remains below a preset threshold, the following parallel processing will be performed: This triggers a recalculation of the semantic evolution path of the term, specifically including re-executing the vector alignment and drift calculation steps in S2 to update the evolution pattern; An early warning message is generated and pushed to the expert review terminal, requesting experts to make an immediate decision on whether the association exists in the knowledge graph based on the current verification results.

[0014] As a preferred embodiment of the medical terminology analysis method based on knowledge graph technology described in this invention, generating quantitative verification results includes: the first update rule being... , ;in, The p-value is obtained from the statistical significance test. The adjustment coefficient is used; the second update rule is... Where c represents the confidence level, The p-value is obtained from the statistical significance test. This is the attenuation coefficient.

[0015] The beneficial effects of this invention are: 1. This invention constructs a sequence of medical terminology knowledge graphs containing time-dimensional information from different versions of medical standard literature. Using a random walk algorithm, word embedding model, and vector space alignment technology, it achieves quantitative calculation of the semantic representation changes of target medical terms at continuous time points, thereby obtaining the semantic evolution path. This process transforms the semantic changes of medical terms from traditional qualitative descriptions to quantifiable analysis, accurately revealing evolutionary patterns such as semantic strengthening, weakening, or subversion, providing a quantitative basis for the adaptive dynamic updating of medical knowledge graphs.

[0016] 2. This invention acquires clinical course data related to target medical terms and constructs a multivariate statistical significance test model based on statistical significance testing methods, achieving clinical quantitative verification of emerging semantics in the semantic evolution path. This mechanism not only ensures the scientific rigor and reliability of the verification process, enabling proactive and accurate perception and verification of emerging semantics, but also transforms the knowledge graph maintenance model from reactive post-event updates to proactive optimization based on prediction and verification, thereby effectively reducing potential medical risks caused by unclear terminology evolution. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 This is a flowchart illustrating the overall process of a medical terminology analysis method based on knowledge graph technology according to the present invention.

[0018] Figure 2 This is a flowchart illustrating the quantitative verification process of a medical terminology analysis method based on knowledge graph technology according to the present invention. Detailed Implementation

[0019] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0020] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0021] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0022] Secondly, the present invention is described in detail with reference to the schematic diagrams. When detailing the embodiments of the present invention, for ease of explanation, the cross-sectional views illustrating the device structure may be partially enlarged, not according to the usual scale. Furthermore, the schematic diagrams are merely examples and should not limit the scope of protection of the present invention. In addition, actual fabrication should include three-dimensional spatial dimensions of length, width, and depth.

[0023] Example 1 Reference Figure 1-2 This is the first embodiment of the present invention, which provides a medical terminology analysis method based on knowledge graph technology, including the following steps: S1: Obtain medical standard documents from different time periods and construct a medical terminology knowledge graph sequence containing time dimension information.

[0024] Specifically, constructing a medical terminology knowledge graph sequence that includes time-related information includes the following steps: Knowledge is extracted from medical literature at each time period to form multiple static knowledge graphs; Each static knowledge graph is assigned a timestamp and organized in chronological order to form a knowledge graph sequence.

[0025] Knowledge extraction refers to the precise location of medical entities in texts through named entity recognition technology, such as gastric cancer, HER2, trastuzumab, positive, negative, etc., and the capture of semantic relationships between entities through relation extraction technology. This transforms the unstructured semantic information in the original medical literature into basic units of a knowledge graph composed of nodes (terms or entities) and edges (semantic relationships), and finally assembles them into a static knowledge graph. Its role is to ensure that the professional knowledge implicit in medical texts can be accurately captured and stored in a structured form, which is convenient for subsequent time comparison and semantic drift calculation.

[0026] A timestamp is a unique time identifier attached to each static knowledge graph generated by knowledge extraction, corresponding to the publication time or version information of the medical document. Specifically, it uses Unix timestamps or ISO 8601 standard time format to record the publication time of the document, and stores this time metadata as an inherent attribute of the graph. This allows each static graph to form a continuously evolving knowledge graph sequence in chronological order. Its function is to clarify the source of semantic differences between different time versions and support the tracking and prediction of subsequent semantic evolution paths, thereby ensuring the interpretability and traceability of dynamic knowledge graph analysis.

[0027] In summary, in step S1, by constructing an independent static knowledge graph for each time version of medical literature and assigning it a timestamp, and then serializing and organizing it into a knowledge graph sequence with time dimension information, the aim is to provide a traceable and quantifiable historical semantic evolution framework for medical terms. Its function is to clearly reveal the evolution path and semantic drift phenomenon of medical concepts such as term definitions and treatment plans over time. Its beneficial effect is to lay an indispensable temporal analysis foundation for subsequent verification of emerging semantics of terms based on real-world clinical data and to achieve adaptive dynamic optimization of the knowledge graph.

[0028] S2: Based on the medical terminology knowledge graph sequence, calculate the semantic representation changes of the target medical term at different points in time to obtain the semantic evolution path of the target medical term.

[0029] Specifically, based on the medical terminology knowledge graph sequence, the semantic representation changes of the target medical term at different points in time are calculated to obtain the semantic evolution path of the target medical term, including the following steps: For each static knowledge graph in the medical terminology knowledge graph sequence, a random walk algorithm is used to generate a node sequence that conforms to the knowledge distribution at the corresponding time point, so as to construct an independent time-series text corpus for each time point; Word embedding models, such as Word2Vec and GloVe, are used to train independently on independent temporal text corpora corresponding to each time point to obtain vector representations of all terms at each time point. By employing vector space alignment technology, term vectors at different time points are mapped to a unified vector space, making term vectors at different time points comparable. For a target medical term, its spatially aligned vector representation at consecutive time points is extracted, and the degree of drift of the term vector between two time points is quantified by calculating the cosine distance or Euclidean distance. By connecting the degree of semantic drift of the target term at all consecutive time points in chronological order, a curve of its semantic similarity over time is plotted. By analyzing the trend of this path, such as steady, rising, falling or drastic fluctuation, the semantic evolution pattern of the term is determined, including but not limited to: semantic stability, semantic strengthening, semantic weakening, semantic subversion or semantic divergence, thus constituting the semantic evolution path of the target medical term.

[0030] The purpose of training word vector models independently at different points in time is to eliminate mutual interference between corpora from different periods, and to ensure that the generated term vector representation can capture and reflect the specific semantic context of the term in the corresponding historical period in an isolated and accurate manner, thereby truly reflecting the era characteristics of semantics.

[0031] Since the models are trained independently, word vectors at different time points reside in different vector spaces, making direct quantitative comparison impossible. Therefore, a vector space alignment technique is introduced. This technique employs orthogonal Procrustes analysis, using the vector space trained on the model at a later time point as the target benchmark. Through linear transformation, word vectors from earlier time points are mapped to this unified vector space, thus ensuring the comparability of term vectors at different time points.

[0032] Furthermore, after obtaining the semantic evolution path of the target medical term, the process also includes prediction and early warning steps: Based on the degree of drift of term vectors, a time series prediction algorithm is used to predict the semantic vector representation of target medical terms at future time points. Based on the prediction results, a semantic evolution trend warning for the term is generated, which includes semantic strengthening, semantic weakening, semantic divergence, or semantic subversion.

[0033] Time series forecasting algorithms, such as ARIMA (Autoregressive Integral Moving Average), LSTM (Long Short-Term Memory), or Prophet, are specifically designed to analyze and chronologically ordered data points to predict future values. Their ability to predict future semantic vectors stems from the fact that the historical vector sequences of terms contain inherent patterns, trends, and laws governing their semantic evolution (such as linear reinforcement, periodic fluctuations, or nonlinear mutations). These algorithms learn and fit these historical patterns to extrapolate the semantic vectors for future time points.

[0034] Preferably, the steps for generating a semantic evolution trend warning are as follows: Based on the predicted future semantic vector, calculate the cosine or Euclidean distance between it and the latest current vector; If the distance is significantly greater than the historical average drift, a semantic subversion warning is generated, indicating that the meaning of the term may undergo a major change, and suggesting that the new meaning represented by the direction of this change is a potential emerging semantic that needs to be verified. If the distance between the predicted vector and the vector cluster representing other semantic concepts decreases significantly, a semantic disambiguation warning is generated, indicating that the term may derive a new semantic branch and suggesting that the new branch is a potential emerging semantic to be verified. If the distance change shows a stable positive or negative trend, a corresponding warning of semantic strengthening or weakening will be generated, indicating that the semantic association between the term and a specific diagnostic concept or entity is strengthening or weakening, and suggesting that the strengthened or weakened association is a potential emerging semantic to be verified.

[0035] The aforementioned early warning generation mechanism maps changes in mathematical distance in vector space into concrete trends of clinical semantic evolution (subversion, misinterpretation, strengthening, or weakening), enabling proactive and accurate classification and early warning of future semantic risks of medical terms. This transforms the maintenance mode of knowledge graphs from reactive post-hoc updates and upgrades to prediction-based pre-preparation and optimization, thereby reducing potential medical risks caused by unclear terminology evolution.

[0036] In summary, step S2 generates a temporal text corpus by applying a random walk algorithm to each static knowledge graph in the medical terminology knowledge graph sequence. A word embedding model is then independently trained on the corpus at different time points to obtain terminology vector representations. Vector space alignment technology is used to achieve cross-temporal vector comparability. Furthermore, the vector distance of the target term at consecutive time points is calculated to quantify its semantic drift, ultimately generating a semantic evolution path. This process transforms the qualitative description of medical terminology semantic changes into quantitative analysis, accurately revealing the strengthening, weakening, or disruptive trends in terminology semantics. It provides reliable temporal evolution evidence for clinical semantic verification and lays a quantitative foundation for adaptive optimization of the knowledge graph. Additionally, it transforms the maintenance mode of the knowledge graph from reactive, passive updates to predictive, proactive preparation and optimization, thereby reducing potential medical risks caused by unclear terminology evolution.

[0037] S3: Obtain clinical course data related to the target medical terminology; calculate the support of the clinical course data for emerging semantics in the semantic evolution path based on statistical significance testing methods, and generate quantitative verification results.

[0038] When acquiring clinical course data related to the target medical terminology, the extracted data is cleaned and standardized, including unifying the coding format (such as using ICD-10 for disease diagnosis coding and ATC for drug coding), processing missing and outlier values, and accurately grouping patients based on the target medical terminology and emerging semantics to form a standardized dataset that can be used for statistical analysis.

[0039] Specifically, based on statistical significance testing methods, the support of clinical course data for emerging semantics in semantic evolution paths is calculated, including: Clinical course data of treatment plans that adopted emerging semantics were extracted as the experimental group, and clinical course data of treatment plans that adopted traditional treatment plans were extracted as the control group. The clinical course data included, but was not limited to, diagnostic information, medication records, surgical procedures, laboratory test results, and efficacy evaluation records. The differences between the experimental group and the control group in pre-defined clinical efficacy indicators were quantitatively calculated. The pre-defined clinical efficacy indicators included at least one of the following: survival rate, incidence of adverse reactions, and symptom relief period. A statistical hypothesis testing model is constructed. The model takes the efficacy index data of the experimental group and the control group as input, calculates and obtains the statistical measure that represents the significance of the difference, and outputs the corresponding p value.

[0040] Furthermore, the differences between the experimental group and the control group in the pre-specified clinical efficacy indicators were quantitatively calculated, including: For the survival rate indicator, the difference is quantified by calculating the log-rank test statistic between the survival curves of the experimental group and the control group; the survival curves are plotted using the Kaplan-Meier method; the log-rank test statistic is used to characterize the difference in survival rate between the experimental group and the control group. For the adverse reaction incidence rate, the difference is calculated by calculating the number of adverse reactions and the total number of adverse reactions in the experimental group and the control group during the observation period, respectively, and the adverse reaction rate of each group is calculated. Then, the ratio or proportion between the two groups is calculated, and the corresponding confidence interval is constructed to quantitatively evaluate the difference in adverse reaction rate between the experimental group and the control group. For the symptom relief period index, the difference was calculated by recording the symptom relief time of subjects in the experimental group and the control group, calculating the difference in the average relief period or the median relief period between the two groups, and evaluating the significance of the difference through statistical tests to quantify the difference in symptom relief period between the two groups.

[0041] Univariate statistical analyses (such as log-rank test, chi-square test, and t-test) were performed on survival rate, adverse reaction rate, and symptom remission period indicators to provide a preliminary assessment of efficacy differences. To control for potential confounding factors and obtain more reliable conclusions, a multivariate statistical significance test model, as described below, was constructed for further analysis.

[0042] Furthermore, constructing a statistical hypothesis testing model includes: To construct a multivariate statistical significance test model to control for the interference of differences in patient baseline characteristics (such as age, gender, disease stage, etc.) on efficacy indicators; The dependent variable of the model is a pre-defined clinical efficacy indicator; The core independent variable of the model is the treatment regimen group (experimental group / control group); Patient age, gender, and disease stage are included as covariates in the model. By fitting the model, the regression coefficients, Wald statistic, and p-values ​​corresponding to the core independent variables are calculated. The p-value is the significance measure of the effect of treatment grouping on efficacy after excluding the interference of covariates.

[0043] Preferably, the multivariate statistical significance test model is a logistic regression model or a Cox proportional hazards model. By selecting the optimal logistic regression model and Cox proportional hazards model, the best statistical test method can be flexibly matched according to different types of clinical efficacy indicators, thereby improving the statistical rationality and explanatory power of the test results, avoiding analytical bias caused by inappropriate model selection, and ultimately enhancing the credibility and clinical applicability of the determination of the difference in efficacy between the experimental group and the control group.

[0044] The preferred mathematical expression for the logistic regression model is: ;in, This is called the logarithmic odds; This represents the predicted probability of a clinical efficacy outcome, a value between 0 and 1. This indicates the grouping of the core independent variable, the treatment regimen. For example, the experimental group is coded as 1, and the control group is coded as 0. , … Represents the covariates introduced into the model; It is the core independent variable The regression coefficients, The odds ratio represents the probability of an event occurring in the experimental group relative to that in the control group. For the intercept term; , , , These are the regression coefficients of the covariates. The process of obtaining the p-value from the model is as follows: The maximum likelihood estimation method is used to fit the model, and the set of coefficient estimates that maximizes the probability of the currently observed therapeutic outcome (e.g., effective / ineffective) is solved iteratively. , , , The algorithm calculates the standard error of the coefficient estimate along with the coefficient estimate, for example: The standard misremembered as Standard error is used to measure the accuracy of regression coefficient estimates; the smaller the standard error, the more reliable the estimate.

[0045] Then, the advantage ratio is calculated: As an effect indicator for quantifying differences in therapeutic efficacy; Then, calculate the Wald statistic: ; Finally, based on the standard normal distribution followed by the Wald statistic, the corresponding p-value is queried or calculated to determine the statistical significance of the difference in treatment efficacy.

[0046] The preferred mathematical expression for the Cox proportional hazards model is: ;in, It refers to the risk function at time t given a covariate X, i.e., the instantaneous mortality risk rate; It is the baseline risk function, representing the risk function when all covariates take the value of 0; , , … The meaning is the same as that of the logistic regression model; It is the core independent variable The regression coefficients, The risk ratio represents the ratio of the risk of death in the experimental group to that in the control group. , , , These are the regression coefficients of the covariates. The process of obtaining the p-value from the model is as follows: Model simulation and output: The partial likelihood estimation method is used to fit the model, and the estimated values ​​of the regression coefficients and their standard errors are obtained iteratively. Next, calculate the risk ratio: As an effect indicator for quantifying differences in survival risk; Then, calculate the Wald statistic: ; Finally, based on the standard normal distribution followed by the Wald statistic, the p-value is queried or calculated to determine the statistical significance of the difference in hazard ratios.

[0047] It should be noted that for logistic regression models, fitting refers to using maximum likelihood estimation to iteratively solve for the set of independent variable coefficients that maximizes the probability of the currently observed therapeutic outcome, thus determining the final form of the model. For Cox proportional hazards models, fitting refers to using partial likelihood estimation to iteratively solve for the hazard ratio and its significance that best characterizes the association between treatment regimen groups and survival time distribution. The fitting process is usually automated by calling standard statistical software, requiring no manual calculation.

[0048] Furthermore, quantitative verification results are generated, including: The preset significance level threshold is ; Compare the calculated p-value with the threshold Compare; If p value < Then, a quantitative verification result of successful clinical validation is generated; If p value ≥ This will generate a quantitative validation result indicating that the clinical validation was unsuccessful.

[0049] Among them, significance level threshold This is a tolerable type of error probability in hypothesis testing, typically set in the fields of medical and life science research. = 0.05 means that there is a 5% probability of incorrectly determining that there is a difference between the new scheme and the traditional scheme, which is an internationally accepted standard.

[0050] By introducing the cleaning and standardization of clinical course data in step S3, and conducting multi-dimensional univariate difference quantification analysis based on the grouping design of the experimental and control groups, followed by correction analysis using multivariate statistical significance test models such as logistic regression or Cox proportional hazards, it is possible to comprehensively evaluate the differences in multiple clinical efficacy indicators such as survival rate, adverse reaction rate, and symptom relief period of emerging semantic corresponding treatment plans, while effectively controlling the potential confounding bias caused by differences in patient baseline characteristics, ensuring the objectivity, scientificity, and statistical credibility of the difference evaluation. Furthermore, by comparing the obtained p-value with the preset significance level, intuitive, quantitative, and verifiable clinical validation results can be generated, thereby achieving objective support and dynamic testing of emerging semantics in the semantic evolution path of target medical terms, providing a solid clinical evidence foundation for the subsequent update and optimization of the knowledge graph.

[0051] S4: Based on the quantitative verification results, dynamically update and optimize the medical terminology knowledge graph sequence.

[0052] Specifically, based on the quantitative verification results, the medical terminology knowledge graph sequence is dynamically updated and optimized, including: If the quantitative validation result indicates successful clinical validation, the first optimization strategy will be automatically executed. The first optimization strategy includes: Based on the p-value of the statistical significance test, the edge weight between the target medical term and the corresponding treatment plan node of the emerging semantic is increased according to the preset first update rule; Add a "clinical evidence support" label to this emerging semantic node in the knowledge graph and record the verification timestamp for traceability and subsequent version management; If the quantitative validation result indicates that clinical validation was unsuccessful, a second optimization strategy will be automatically executed. The second optimization strategy includes: Based on the p-value of the statistical significance test, the association confidence between the target medical term and the treatment plan node corresponding to the emerging semantics is reduced according to the preset second update rule. If the association confidence level remains below a preset threshold for an extended period (e.g., for N consecutive validation cycles), the following parallel processing will be performed: This triggers a recalculation of the semantic evolution path of the term, specifically including re-executing the vector alignment and drift calculation steps in S2 to update the evolution pattern; An early warning message is generated and pushed to the expert review terminal, requesting experts to make an immediate decision on whether the association exists in the knowledge graph based on the current verification results.

[0053] The preset threshold and duration can be set by domain experts based on the actual application scenario. Typically, the threshold is set to 0.2. A lower threshold (such as 0.1) means that the system is more lenient and requires more verification failures before triggering a recalculation; a higher threshold (such as 0.3) means that the system is more sensitive and can react to unreliable associations more quickly.

[0054] The warning information includes at least the following: Core warning theme: Warning regarding the failure of clinical validation of the association between target medical terms and emerging semantics; Identity information: Unique identifiers for target medical terms, emerging semantics, and their corresponding treatment plans; Validation results data: p-values ​​and confidence intervals for statistical significance tests in this and historical data; System status data: current association confidence level and its historical changes; Decision-making background information: Visual charts pointing to the semantic evolution path, for experts to refer to the origin and trend of semantic drift; Suggested action: Options for experts to choose from, such as "Confirm and disable association", "Ignore this alert", or "Mark as under observation".

[0055] It should be noted that by defining the early warning information as a structured report containing a complete chain of evidence, a panoramic data view and an actionable workflow can be provided for expert decision-making, ensuring the accuracy, efficiency and traceability of human-machine collaborative decision-making, thereby ultimately achieving a deep integration of automation and human expert supervision in the knowledge graph maintenance process.

[0056] Preferred, the first update rule is: , ; in, The p-value is obtained from the statistical significance test. This is the adjustment coefficient.

[0057] The preferred second update rule is: ; Where c represents the confidence level. The p-value is obtained from the statistical significance test. This is the attenuation coefficient.

[0058] It should also be noted that the weighting adjustment coefficient With attenuation coefficient This can be trained based on historical data, or pre-set by domain experts according to the magnitude of the association weights and confidence levels in the knowledge graph. For example, It can be set to a value between 0.1 and 1.0 to ensure that the weight update range is appropriate; It can be set to a value between 1.0 and 5.0 to control the confidence level to drop rapidly below the threshold after several failed verifications.

[0059] Step S4 above automatically triggers a differentiated optimization strategy based on the quantitative verification results (p-value). If successful, the edge weights of the knowledge graph are increased according to preset rules and clinical evidence annotations are added. If unsuccessful, the association confidence is reduced or the semantic path is recalculated. This achieves closed-loop, adaptive, and evidence-driven dynamic updates of the medical terminology knowledge graph, thereby significantly improving the timeliness, accuracy, and reliability of the knowledge base. At the same time, by generating early warning information, the structured report of the conclusion that failed clinical verification and its complete evidence chain is pushed to the expert review terminal, effectively reducing the risk of clinical decision-making caused by knowledge lag or errors.

[0060] In summary, this invention constructs a sequence of medical terminology knowledge graphs containing time-dimensional information from different versions of medical standard literature. By employing random walk algorithms, word embedding models, and vector space alignment techniques, it achieves quantitative calculation of the semantic representation changes of target medical terms at continuous time points, thereby obtaining the semantic evolution path. This process transforms the semantic evolution of medical terms from traditional qualitative descriptions to quantifiable analysis, accurately revealing evolutionary patterns such as semantic strengthening, weakening, or subversion, providing a quantitative basis for the adaptive dynamic updating of medical knowledge graphs. Furthermore, this invention acquires clinical course data related to target medical terms and constructs a multivariate statistical significance test model based on statistical significance testing methods, achieving clinical quantitative verification of emerging semantics in the semantic evolution path. This mechanism not only ensures the scientific rigor and reliability of the verification process, enabling proactive and accurate perception and verification of emerging semantics, but also transforms the maintenance mode of the knowledge graph from reactive post-updates to proactive optimization based on prediction and verification, thereby effectively reducing potential medical risks caused by unclear terminology evolution.

[0061] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for analyzing medical terminology based on knowledge graph technology, characterized in that, include: Obtain medical standard documents from different time periods and construct a sequence of medical terminology knowledge graphs containing time-related information. Based on the medical terminology knowledge graph sequence, the semantic representation changes of the target medical term at different points in time are calculated to obtain the semantic evolution path of the target medical term. Acquire clinical course data related to the target medical term; calculate the support of the clinical course data for emerging semantics in the semantic evolution path based on statistical significance testing methods, and generate quantitative verification results; Based on the quantitative verification results, the medical terminology knowledge graph sequence is dynamically updated and optimized.

2. The medical terminology analysis method based on knowledge graph technology as described in claim 1, characterized in that: Constructing a medical terminology knowledge graph sequence that includes time-related information includes the following steps: Knowledge is extracted from medical literature at each time period to form multiple static knowledge graphs; Each static knowledge graph is assigned a timestamp and organized in chronological order to form the knowledge graph sequence.

3. The medical terminology analysis method based on knowledge graph technology as described in claim 2, characterized in that: Based on the medical terminology knowledge graph sequence, the semantic representation changes of the target medical term at different points in time are calculated to obtain the semantic evolution path of the target medical term, including the following steps: For each of the static knowledge graphs in the medical terminology knowledge graph sequence, a node sequence conforming to the knowledge distribution at the corresponding time point is generated by a random walk algorithm, so as to construct an independent time-series text corpus for each time point; A word embedding model is used, which is trained independently on an independent temporal text corpus corresponding to each time point to obtain the vector representation of all terms at each time point; By employing vector space alignment technology, term vectors at different time points are mapped to a unified vector space, making term vectors at different time points comparable. For a target medical term, its spatially aligned vector representation at consecutive time points is extracted, and the degree of drift of the term vector between two time points is quantified by calculating the cosine distance or Euclidean distance. By connecting the semantic drift of the target term at all consecutive time points in chronological order, a curve showing the change of its semantic similarity over time is plotted. By analyzing the trend of this path, the semantic evolution pattern of the term is determined, thus constructing the semantic evolution path of the target medical term.

4. The medical terminology analysis method based on knowledge graph technology as described in claim 3, characterized in that: After obtaining the semantic evolution path of the target medical term, the process also includes prediction and early warning steps: Based on the degree of drift of the term vector, a time series prediction algorithm is used to predict the semantic vector representation of the target medical term at future time points; Based on the prediction results, a semantic evolution trend warning for the term is generated, which includes semantic strengthening, semantic weakening, semantic divergence, or semantic subversion.

5. The medical terminology analysis method based on knowledge graph technology as described in claim 4, characterized in that: The steps for generating the semantic evolution trend warning are as follows: Based on the predicted future semantic vector, calculate the cosine or Euclidean distance between it and the latest current vector; If the distance is significantly greater than the historical average drift, a semantic subversion warning is generated, indicating that the meaning of the term may undergo a major change, and suggesting that the new meaning represented by the direction of this change is a potential emerging semantic that needs to be verified. If the distance between the predicted vector and the vector cluster representing other semantic concepts decreases significantly, a semantic disambiguation warning is generated, indicating that the term may derive a new semantic branch and suggesting that the new branch is a potential emerging semantic to be verified. If the distance change shows a stable positive or negative trend, a corresponding warning of semantic strengthening or weakening will be generated, indicating that the semantic association between the term and a specific diagnostic concept or entity is strengthening or weakening, and suggesting that the strengthened or weakened association is a potential emerging semantic to be verified.

6. The medical terminology analysis method based on knowledge graph technology as described in claim 5, characterized in that: Based on statistical significance testing, the support of clinical course data for emerging semantics in the semantic evolution path is calculated, including: Clinical course data of treatment plans that adopted the new semantics were extracted as the experimental group, and clinical course data of traditional treatment plans were extracted as the control group; wherein, the clinical course data includes, but is not limited to, diagnostic information, medication records, surgical procedures, laboratory test results and efficacy evaluation records; The differences between the experimental group and the control group in terms of preset clinical efficacy indicators are quantitatively calculated; the preset clinical efficacy indicators include at least one of survival rate, adverse reaction rate, and symptom relief period. A statistical hypothesis testing model is constructed. The model takes the efficacy index data of the experimental group and the control group as input, calculates and obtains the statistical measure representing the significance of the difference, and outputs the corresponding p value.

7. The medical terminology analysis method based on knowledge graph technology as described in claim 6, characterized in that: Constructing the statistical hypothesis testing model includes: A multivariate statistical significance test model was constructed to control for the interference of patient baseline characteristics differences on efficacy indicators; the multivariate statistical significance test model was a logistic regression model or a Cox proportional hazards model. The dependent variable of the model is a preset clinical efficacy indicator; The core independent variable of the model is the treatment regimen grouping; Patient age, gender, and disease stage are included as covariates in the model. By fitting the model, the regression coefficients, Wald statistic, and p-values ​​corresponding to the core independent variables are calculated. The p-value is the significance measure of the effect of treatment grouping on efficacy after excluding the interference of covariates.

8. The medical terminology analysis method based on knowledge graph technology as described in claim 7, characterized in that: Generate quantitative verification results, including: The preset significance level threshold is ; Compare the calculated p-value with the threshold Compare; If p value < Then, a quantitative verification result of successful clinical validation is generated; If p value ≥ This will generate a quantitative validation result indicating that the clinical validation was unsuccessful.

9. The medical terminology analysis method based on knowledge graph technology as described in claim 1, characterized in that: Based on the quantitative verification results, the medical terminology knowledge graph sequence is dynamically updated and optimized, including: If the quantitative verification result indicates successful clinical validation, then the first optimization strategy is automatically executed, which includes: Based on the p-value of the statistical significance test, the edge weight between the target medical term and the treatment plan node corresponding to the emerging semantic is increased according to the preset first update rule. Add clinical evidence support annotations to this emerging semantic node in the knowledge graph and record the verification timestamp for traceability and subsequent version management; If the quantitative verification result indicates that the clinical verification was unsuccessful, a second optimization strategy will be automatically executed. The second optimization strategy includes: Based on the p-value of the statistical significance test, the association confidence between the target medical term and the treatment plan node corresponding to the emerging semantics is reduced according to the preset second update rule. If the association confidence level remains below a preset threshold, the following parallel processing will be performed: This triggers a recalculation of the semantic evolution path of the term, specifically including re-executing the vector alignment and drift calculation steps in S2 to update the evolution pattern; An early warning message is generated and pushed to the expert review terminal, requesting experts to make an immediate decision on whether the association exists in the knowledge graph based on the current verification results.

10. The medical terminology analysis method based on knowledge graph technology as described in claim 9, characterized in that: The first update rule is , ;in, The p-value is obtained from the statistical significance test. The adjustment coefficient is used; the second update rule is... Where c represents the confidence level, The p-value is obtained from the statistical significance test. This is the attenuation coefficient.

Images