Knowledge graph construction system and method based on orthopedic clinic

Abstract

The invention belongs to the technical field of medical information interaction, and provides a knowledge graph construction system and method based on orthopaedic clinic, and the method comprises the steps: in an emergency treatment construction scene, combining emergency treatment terms in emergency treatment data in a previous period with professional terms corresponding to a data storage library, obtaining corresponding term mapping values, and carrying out the calculation of the corresponding term mapping values; therefore, the stability degree of the mapping relation between the emergency treatment terms and the professional terms in the time dimension is reflected through the corresponding term mapping values, so that the problem of low mapping corresponding accuracy in the term mapping process due to diversity and complexity of the emergency treatment terms is solved; therefore, the problem of low accuracy of the subsequently constructed knowledge graph is solved, a doctor can make a more accurate decision in the emergency clinical treatment, and the treatment effect of the emergency clinical treatment is improved.

Description

Technical Field

[0001] The present invention belongs to the field of medical information interaction technology, and specifically relates to a knowledge graph construction system and method based on orthopedic clinical practice. Background Art

[0002] In the process of orthopedic clinical diagnosis and treatment, knowledge graphs, as a structured knowledge representation method, can effectively integrate, associate and display knowledge in the field of orthopedics, and provide doctors with comprehensive decision-making support. Especially in emergency orthopedic clinical scenarios, due to the urgency of the patient's condition, doctors need to make accurate diagnosis and treatment decisions quickly. Therefore, it is particularly important to construct orthopedic clinical knowledge graphs in real time and accurately.

[0003] In existing technologies, the construction of orthopedic clinical knowledge graphs often lacks a clear definition and feature extraction of emergency orthopedic data. This results in a lack of quantitative basis and feature benchmarks when identifying emergency scenarios. This makes it difficult for the system to quickly and accurately identify and process relevant data in emergency situations, thus affecting the real-time construction efficiency of the knowledge graph. Moreover, the terminology involved in emergency orthopedic clinical practice is diverse and complex. Existing technologies often suffer from low mapping accuracy during the term mapping process. This is mainly due to the diversity and complexity of emergency terminology and the instability of the mapping relationship between terms. Therefore, inaccurate term mapping will directly affect the accuracy of the knowledge graph, and then affect the accuracy and real-time performance of the knowledge graph, and fail to provide timely decision support for doctors.

[0004] To this end, the present invention provides a knowledge graph construction system and method based on orthopedic clinical practice. Summary of the Invention

[0005] In order to make up for the deficiencies of the prior art, at least one technical problem raised in the background technology is solved.

[0006] The technical solution adopted by the present invention to solve its technical problem is: First, a method for constructing a knowledge graph based on orthopedic clinical practice includes: In previous cycles, the orthopedic data chain in the data repository was extracted and analyzed to obtain the range of emergency characteristics; During the current test cycle, the orthopedic data chain currently submitted to the data repository is analyzed in the time dimension and compared with the emergency feature range to determine whether the current construction scenario is an emergency construction scenario.

[0007] If a scenario is built for emergency care, an emergency terminology mapping model is constructed by combining the emergency terminology in the emergency data of previous cycles with the corresponding professional terminology in the data repository.

[0008] During the current testing cycle, the emergency terminology mapping model is subjected to emergency term recognition diagnosis, the recognition capability of the emergency terminology mapping model is evaluated, and an orthopedic clinical knowledge graph is constructed in the emergency construction scenario.

[0009] As a further solution of the present invention, the orthopedic data chain in the data repository is extracted and analyzed, and the process is as follows: Randomly select an orthopedic data link in the data repository, obtain the entry time of adjacent orthopedic data in the orthopedic data link, perform difference processing, and take the absolute value to output the adjacent entry time interval. If the adjacent entry time interval is less than or equal to the preset adjacent entry time interval, it is an emergency orthopedic data link.

[0010] As a further solution of the present invention, the process of obtaining the emergency feature range is as follows: Obtain the process submission time of each emergency orthopedic data link, compare them, and extract the maximum and minimum process submission time of the emergency orthopedic data link respectively.

[0011] The adjacent entry time intervals corresponding to each emergency orthopedic data link are compared, and the maximum adjacent entry time interval corresponding to each emergency orthopedic data link is extracted respectively. All the maximum adjacent entry time intervals are compared to obtain the relative minimum adjacent entry time interval and the relative maximum adjacent entry time interval.

[0012] The maximum value of the process submission time and the relative maximum adjacent entry time interval are summed, as well as the minimum value of the process submission time and the relative minimum adjacent entry time interval are summed to obtain the maximum value and the minimum value of the emergency feature, and the emergency feature range is obtained.

[0013] A further solution of the present invention is to divide the current test cycle into several test time periods, obtain the proportion of the duration of the entire orthopedic data link submission process in each test time period to the duration corresponding to the test time period, and obtain the process submission duration ratio.

[0014] In each test period, the entry time of adjacent orthopedic data in the orthopedic data chain is extracted, difference processing is performed, and the absolute value is taken to output the current adjacent entry time interval.

[0015] All current adjacent entry time intervals in the orthopedic data chain are averaged and the average entry time interval of the current process is output.

[0016] The process submission time ratio is summed with the average of the current process entry time intervals to obtain the scenario evaluation value. If the scenario evaluation value is within the emergency feature range, the corresponding construction scenario is marked as an emergency construction scenario.

[0017] A further solution of the present invention is to divide the previous period into several previous time periods, and within the previous time period, arbitrarily extract an emergency term from an emergency data in the emergency orthopedic data chain as the target term, obtain the professional term corresponding to the target term in the data repository within the previous period, and combine them according to the time series to obtain multiple groups of term correspondence groups.

[0018] Among multiple terminology correspondence groups, extract the terminology correspondence groups that have the same professional terminology, and obtain the proportion of the total number of all terminology correspondence groups to obtain the corresponding number ratio of the same professional terminology.

[0019] As a further solution of the present invention, based on the corresponding group of terms with the same professional terminology, all corresponding previous time periods are obtained as the same professional terminology time periods, and are sorted according to the time series to obtain multiple same professional terminology time period sequences, and the sorting of all adjacent same professional terminology time periods in each same professional terminology time period sequence in the previous cycle is input into the convolutional layer in the convolutional neural network to obtain multiple training convolutional layers.

[0020] The order of adjacent periods of the same professional terminology in multiple training convolutional layers in previous cycles is input into the improved Euclidean distance calculation formula, and the output is the corresponding stable value of the term.

[0021] As a further solution of the present invention, the construction process of the emergency terminology mapping model is as follows: The geometric product of the corresponding quantity ratio of the same professional term and the corresponding stable value of the term is calculated, and the corresponding term mapping value is output.

[0022] If the corresponding term mapping value is greater than or equal to the corresponding term mapping threshold, a mapping stability signal is generated, and an emergency term mapping model is constructed based on multiple training convolutional layers corresponding to the generated mapping stability signals.

[0023] As a further solution of the present invention, within any test period, the mapping duration between any test emergency term and the test professional term is obtained, the proportion of the total duration of the test period is calculated, and the averaging processing is performed to output the period mapping duration ratio.

[0024] The time period mapping duration ratios corresponding to all test periods in the current test cycle are averaged and the mapping duration test value is output.

[0025] The mapping duration ratios within each test period are sorted according to the corresponding test time to construct a period test sequence. The period test sequences corresponding to adjacent test periods are input into the Pearson distance formula, and then averaged and calculated to obtain the stable value of the mapping duration.

[0026] As a further solution of the present invention, if the recognition diagnosis value is less than or equal to the recognition diagnosis threshold, a stable and efficient signal is generated, the emergency term mapping model is imported into the data acquisition system, and the emergency knowledge graph is constructed according to the emergency term mapping model.

[0027] If the recognition diagnosis value is greater than the recognition diagnosis threshold, a fluctuating inefficient signal is generated, the emergency term mapping model is adjusted, and the adjusted emergency term mapping model is re-identified and diagnosed until a stable and efficient signal is generated, imported into the data acquisition system, and the emergency knowledge graph is constructed according to the emergency term mapping model.

[0028] Secondly, a knowledge graph construction system based on orthopedic clinical practice includes the following modules: Feature recognition module: In the previous cycle, the orthopedic data chain in the data repository was extracted and analyzed to obtain the emergency feature range.

[0029] Scenario determination module: During the current test cycle, the orthopedic data chain currently submitted to the data repository is analyzed in the time dimension and compared with the emergency feature range to determine whether the current construction scenario is an emergency construction scenario.

[0030] Model building module: If the scenario is built for emergency care, the emergency terminology in the emergency data in previous cycles and the corresponding professional terminology in the data repository are combined to build an emergency term mapping model.

[0031] Graph construction module: During the current testing cycle, the emergency terminology mapping model is used to perform emergency terminology recognition diagnosis, evaluate the recognition capability of the emergency terminology mapping model, and construct an orthopedic clinical knowledge graph in the emergency construction scenario.

[0032] The beneficial effects of the present invention are as follows: 1. In the previous cycle, the present invention extracted and analyzed the orthopedic data chain in the data repository to obtain the emergency feature range, thereby using the emergency feature range as the key attribute to define the emergency orthopedic data, and providing a quantitative basis and feature benchmark for subsequent emergency scene identification and knowledge graph construction. In the current test cycle, the orthopedic data chain currently submitted to the data repository is analyzed in the time dimension, and combined with the emergency feature range for comparison to obtain a scenario evaluation value, thereby reflecting the time efficiency of the orthopedic data chain from the beginning to the completion of the entire diagnostic process, and reflecting the time compactness of each link in the orthopedic data chain, thereby providing data support for the subsequent real-time generation of the emergency knowledge graph and embedding it into the electronic medical record system, thereby improving the efficiency of emergency clinical decision-making.

[0033] 2. If the present invention is an emergency construction scenario, the emergency terms in the emergency data in the previous period and the corresponding professional terms in the data repository are combined to obtain the corresponding term mapping values, so that the corresponding term mapping values ​​reflect the stability of the mapping relationship between the emergency terms and the professional terms in the time dimension, thereby solving the problem of low mapping accuracy during the term mapping process due to the diversity and complexity of emergency terms, and then causing the problem of low accuracy of the knowledge graph constructed subsequently, which helps doctors make more accurate decisions in emergency clinics and improves the treatment effect of emergency clinics.

[0034] 3. During the current test cycle, the present invention performs emergency terminology recognition diagnosis on the emergency terminology mapping model, evaluates the recognition ability of the emergency terminology mapping model, and obtains the recognition diagnosis value, thereby reflecting the recognition ability of the emergency terminology mapping model in the timeliness dimension through the recognition diagnosis value. Based on the recognition ability results, the emergency terminology mapping model is optimized and adjusted in a targeted manner to ensure that the emergency terminology mapping model can perform mapping work in a timely and stable manner in the emergency construction scenario, thereby improving the accuracy of the orthopedic clinical knowledge graph after construction, providing timely decision support for doctors in the emergency construction scenario, and improving the treatment effect of emergency clinical medicine. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] The present invention will be further described below with reference to the accompanying drawings.

[0036] Figure 1 This is a flowchart of the steps of a method for constructing a knowledge graph based on orthopedic clinical practice in the present invention.

[0037] Figure 2 This is a schematic diagram of a knowledge graph construction system based on orthopedic clinical practice in the present invention. DETAILED DESCRIPTION

[0038] In order to make the technical means, creative features, objectives and effects achieved by the present invention easier to understand, the present invention is further described below in conjunction with specific implementation methods.

[0039] Example 1 See also Figure 1 As shown, the embodiment of the present invention describes a method for constructing a knowledge graph based on orthopedic clinical practice. Since the scenario for constructing a knowledge graph for orthopedic clinical practice is not single, there are generally emergency orthopedic clinical scenarios and non-emergency orthopedic clinical scenarios. However, in the emergency orthopedic clinical scenario, due to the need to provide timely treatment to patients, there will be some links in the construction of the knowledge graph that cannot be submitted and uploaded in time and need to be supplemented later, which will affect the patient's treatment decision. The method includes the following steps: Step 1: In the previous cycle, the orthopedic data chain in the data repository was extracted and analyzed to obtain the emergency feature range.

[0040] It should be noted that the orthopedic data chain stored in the data repository is all orthopedic data of each orthopedic clinical diagnostic process, among which all data of each diagnostic process include orthopedic diagnosis and treatment data, orthopedic test result data and orthopedic treatment data.

[0041] Those skilled in the art will understand that orthopedic diagnosis and treatment data include: physical diagnosis and treatment data, differential diagnosis and treatment data, and diagnosis and treatment decision data.

[0042] Orthopedic test result data includes: imaging test result data, laboratory test result data and pathological test result data.

[0043] Orthopedic treatment time includes: non-surgical treatment data and surgical treatment data.

[0044] In some embodiments, an orthopedic data chain in a data repository is arbitrarily selected, the entry time of adjacent orthopedic data in the orthopedic data chain is obtained, difference processing is performed, and the absolute value is taken to output the adjacent entry time interval.

[0045] Compare the adjacent recording time intervals with the preset adjacent recording time intervals. The process is as follows: If the adjacent entry time interval is greater than the preset adjacent entry time interval, it means that the entry time interval of adjacent orthopedic data in the analyzed orthopedic data chain is long, the entry is not urgent, and it is not an emergency orthopedic data chain.

[0046] If the adjacent entry time interval is less than or equal to the preset adjacent entry time interval, it means that the entry time interval of adjacent orthopedic data in the analyzed orthopedic data chain is long, the entry is more urgent, and it is an emergency orthopedic data chain.

[0047] Obtain the process submission time of each emergency orthopedic data link, compare them, and extract the maximum and minimum process submission time of the emergency orthopedic data link respectively.

[0048] The adjacent entry time intervals corresponding to each emergency orthopedic data link are compared, and the maximum adjacent entry time interval corresponding to each emergency orthopedic data link is extracted respectively.

[0049] The maximum adjacent entry time intervals corresponding to each emergency orthopedic data link are compared, and the minimum and maximum values ​​within all maximum adjacent entry time intervals are extracted as the relative minimum adjacent entry time interval and the relative maximum adjacent entry time interval.

[0050] The maximum process submission duration and the relative maximum adjacent entry time interval are summed to obtain the maximum value of the emergency feature.

[0051] The minimum value of the process submission duration and the relatively minimum adjacent entry time interval are summed to output the minimum value of the emergency feature.

[0052] The emergency feature range is constructed based on the maximum value and the minimum value of the emergency feature.

[0053] It should be noted that the meaning of the emergency feature range is: defining the key attributes of emergency orthopedic data from the time dimension and process feature dimension. From the time dimension, one of the core characteristics of the emergency scene is the time urgency of the diagnosis and treatment process. Emergency patients usually require rapid diagnosis and treatment. The entry time interval of each link in the data chain (such as physical examination, imaging detection, treatment decision-making, etc.) is relatively short. From the process dimension, the diagnosis and treatment process of emergency orthopedics has strong coherence and timeliness. In the emergency data chain, the entry time of key links (such as imaging detection, surgical treatment) needs to be closely connected. Therefore, it provides a quantitative basis and feature benchmark for subsequent emergency scene recognition and knowledge graph construction.

[0054] Step 2: During the current test cycle, the orthopedic data chain currently submitted to the data repository is analyzed in terms of time dimension, and compared with the emergency feature range to determine whether the current construction scenario is an emergency construction scenario.

[0055] In some embodiments, the current test cycle is divided into several test periods.

[0056] It should be noted that the time interval corresponding to each test period is equal, and each test period corresponds to one orthopedic data link process submission.

[0057] During the test period, the duration of the entire orthopedic data link submission process is obtained, and the ratio is calculated with the duration corresponding to the test period to obtain the process submission duration ratio.

[0058] During the test period, the entry time of adjacent orthopedic data in the orthopedic data chain is extracted, difference processing is performed, and the absolute value is taken to output the current adjacent entry time interval.

[0059] All current adjacent entry time intervals in the orthopedic data chain are averaged and the average entry time interval of the current process is output.

[0060] The process submission duration ratio is summed with the average of the current process entry intervals, and the output is the scenario evaluation value.

[0061] It can be understood that the meaning of the scenario evaluation value is: by combining the comprehensive indicators of time urgency and process continuity, it is used to quantify the emergency feature matching degree of the current orthopedic data chain. On the one hand, the process submission time ratio reflects the time efficiency of the orthopedic data chain from the beginning to the completion of the entire diagnostic process. On the other hand, the current process entry time interval reflects the time compactness of each link in the orthopedic data chain (such as consultation, testing, and treatment decision-making). Therefore, non-emergency data is filtered through time indicators, and emergency-related nodes in the orthopedic data chain are extracted, providing data support for the subsequent real-time generation of emergency knowledge graphs embedded in the electronic medical record system, improving the efficiency of emergency clinical decision-making, and automatically screening non-emergency data, reducing the workload of manual labeling, and reducing the ineffective consumption of storage and computing resources.

[0062] The scenario evaluation values ​​are compared with the emergency feature ranges as follows: If the scenario evaluation value exists within the emergency feature range, it means that the time period for generating the entire current orthopedic data chain is short, and the time intervals for submission of each link in the current orthopedic data chain are relatively tight, an emergency submission signal is generated, and the corresponding construction scenario is marked as an emergency construction scenario.

[0063] If the scenario evaluation does not exist within the emergency feature range, it means that the time period for generating the entire current orthopedic data chain is long, and the time intervals for submission of each link in the current orthopedic data chain are less urgent, a non-emergency submission signal is generated, and the corresponding construction scenario is marked as a non-emergency construction scenario.

[0064] The technical solution of the embodiment of the present invention is: in the previous cycle, the orthopedic data chain in the data repository is extracted and analyzed to obtain the emergency feature range, so as to use the emergency feature range as the key attribute to define the emergency orthopedic data, and provide a quantitative basis and feature benchmark for subsequent emergency scene recognition and knowledge graph construction. In the current test cycle, the orthopedic data chain currently submitted to the data repository is analyzed in the time dimension, and combined with the emergency feature range for comparison to obtain a scenario evaluation value, so as to reflect the time efficiency of the orthopedic data chain from the beginning to the completion of the entire diagnostic process, and reflect the time compactness of each link in the orthopedic data chain through the scenario evaluation value, thereby providing data support for the subsequent real-time generation of the emergency knowledge graph and embedding it into the electronic medical record system, thereby improving the efficiency of emergency clinical decision-making.

[0065] Example 2 See also Figure 1 As shown, the method for constructing a knowledge graph based on orthopedic clinical practice according to an embodiment of the present invention includes the following steps: Step 3: If the scenario is built for emergency care, an emergency terminology mapping model is constructed based on the emergency terminology in the emergency data of previous cycles and the corresponding professional terminology in the data repository.

[0066] In some embodiments, the past cycle is divided into a number of past time periods.

[0067] It should be noted that the duration interval of each previous period is equal, and each previous period corresponds to a process submission of the emergency orthopedics data link.

[0068] In the past period, an emergency term in the emergency data was randomly extracted from the emergency orthopedics data chain as the target term.

[0069] Based on the target term, the corresponding professional terms in the data repository in the previous period are obtained and combined according to the time series to obtain multiple groups of term correspondence groups.

[0070] Among multiple terminology correspondence groups, extract the terminology correspondence groups that have the same professional terminology, and calculate the proportion of the total number of all terminology correspondence groups to obtain the number ratio of the same professional terminology.

[0071] Based on the term corresponding groups with the same professional term, all previous time periods of the corresponding appearance are obtained as the same professional term time periods, and are sorted according to the time series to obtain multiple same professional term time period sequences.

[0072] In any sequence of periods with the same professional terminology, the order of all adjacent periods with the same professional terminology in previous cycles is input into the convolutional layer of the convolutional neural network to obtain multiple training convolutional layers.

[0073] For example, there are five same-professional-terminology periods in the same-professional-terminology period sequence, namely, A same-professional-terminology period, B same-professional-terminology period, C same-professional-terminology period, D same-professional-terminology period and E same-professional-terminology period.

[0074] Adjacent periods with the same professional terminology are input into the convolutional layer of the convolutional neural network to obtain a training convolutional layer, wherein the training convolutional layer includes period A with the same professional terminology and period B with the same professional terminology, the training convolutional layer includes period B with the same professional terminology and period C with the same professional terminology, the training convolutional layer includes period C with the same professional terminology and period D with the same professional terminology, and the training convolutional layer includes period D with the same professional terminology and period E with the same professional terminology.

[0075] It should be noted that the purpose of using convolutional neural networks is to identify the occurrence patterns of emergency terms in time series. By inputting adjacent time periods of the same professional terms into the convolutional layer of the convolutional neural network, the frequency of use of these terms in the time dimension can be analyzed, and which terms appear frequently in time can be identified, which helps to reduce errors in term mapping and improve the accuracy of knowledge graph construction.

[0076] The order of adjacent terms of the same professional terminology in multiple training convolutional layers in the previous cycle is input into the improved Euclidean distance calculation formula, and the output is the corresponding stable value of the term. ; Specifically: The Euclidean distance calculation formula is: ,in, Denoted as the total number of training convolutional layers, Expressed as The first The ranking of the same professional term period in the previous cycle, Expressed as The first The ranking of the same professional term time period in previous cycles.

[0077] It should be noted that the purpose of using the Euclidean distance calculation formula is: by calculating the corresponding stable values ​​of the terms, those terms with more stable usage can be identified and mapped first. Therefore, on the basis of improving the accuracy of term mapping, the accuracy of the knowledge graph constructed subsequently is improved.

[0078] The geometric product of the corresponding quantity ratio of the same professional term and the corresponding stable value of the term is calculated, and the corresponding term mapping value is output.

[0079] Those skilled in the art can understand that the meaning represented by the corresponding term mapping value is: by combining the corresponding quantity ratio of the same professional terms and the corresponding stability value of the terms, on the one hand, the quantity ratio of the same professional terms reflects the quantitative proportion of the mapping relationship between emergency terms and professional terms; on the other hand, the corresponding stability value of the terms reflects the stability of the mapping relationship between emergency terms and professional terms in the time dimension, thereby solving the problem of low mapping correspondence accuracy in the term mapping process due to the diversity and complexity of emergency terms, and then causing the problem of low accuracy of the knowledge graph constructed subsequently, which helps doctors make more accurate decisions in emergency clinics and improve the treatment effect of emergency clinics.

[0080] Compare the corresponding term mapping value with the corresponding term mapping threshold as follows: If the corresponding term mapping value is greater than or equal to the corresponding term mapping threshold, it means that the mapping relationship between the emergency term and the professional term has a large proportion and is relatively stable in the time dimension, generating a mapping stability signal.

[0081] If the corresponding term mapping value is less than the corresponding term mapping threshold, it means that the mapping relationship between the emergency term and the professional term has a small proportion and is relatively unstable in the time dimension, generating a mapping instability signal.

[0082] Based on multiple training convolutional layers corresponding to the generated mapping stability signals, an emergency term mapping model is constructed; The specific implementation scheme of the present invention is: if the emergency scenario is constructed, the emergency terms in the emergency data in the previous cycle and the corresponding professional terms in the data repository are combined to obtain the corresponding term mapping values, so that the corresponding term mapping values ​​reflect the stability of the mapping relationship between the emergency terms and the professional terms in the time dimension, thereby solving the problem of low mapping accuracy during the term mapping process due to the diversity and complexity of emergency terms, and then causing the problem of low accuracy of the knowledge graph constructed subsequently, which helps doctors make more accurate decisions in emergency clinics and improve the treatment effect of emergency clinics.

[0083] Example 3 See also Figure 1 As shown, the method for constructing a knowledge graph based on orthopedic clinical practice according to an embodiment of the present invention includes the following steps: Step 4: During the current test cycle, perform emergency terminology recognition diagnosis on the emergency terminology mapping model, evaluate the recognition ability of the emergency terminology mapping model, and construct an orthopedic clinical knowledge graph in the emergency construction scenario.

[0084] In some embodiments, during the current test cycle, the test emergency terms in all test periods are input into the emergency term mapping model, and the test professional terms mapped to each test emergency term in each test period are counted and a diagnostic comparison is performed. The process is as follows: In any test period, the mapping duration between any test emergency term and any test professional term is obtained, and the ratio is calculated with the total duration of the test period to output the mapping duration ratio.

[0085] All mapping duration ratios within the test period are averaged and the mapping duration ratio of the period is output.

[0086] The time period mapping duration ratios corresponding to all test periods in the current test cycle are averaged and the mapping duration test value is output.

[0087] The mapping duration ratios within each test period are sorted according to the corresponding test time to construct a period test sequence.

[0088] It should be noted that the total number of mapping duration ratios within the test sequence in each period is the same; In the current test cycle, the test sequences corresponding to the adjacent test periods are input into the Pearson distance formula, and the output is the stable value of the adjacent time length. .

[0089] Specifically, the Pearson distance formula: , calculate the stable value of adjacent time length ,in, Indicates the first test period in the test sequence corresponding to one of the adjacent test periods. The mapping duration ratio, and It is expressed as the mean of the mapping duration ratios in all test sequences corresponding to one of the test periods in adjacent test periods. Indicates the first test period in the test sequence corresponding to another test period in the adjacent test period. The mapping duration ratio, and It is expressed as the mean of the mapping duration ratios in all test sequences corresponding to another test period in the adjacent test period. It is expressed as the total number of mapping duration ratios in all period test sequences corresponding to one of the test periods in adjacent test periods.

[0090] The adjacent duration stability values ​​corresponding to all adjacent test periods are averaged and the mapped duration stability value is output.

[0091] The mapping duration stability value and the mapping duration test value are input into the geometric product method, and the identification diagnosis value is output.

[0092] People skilled in the art can understand that the meaning of the recognition diagnosis value is: it is used to evaluate the recognition ability of the emergency term mapping model. On the one hand, the mapping time test value reflects the mapping efficiency of the emergency term mapping model when mapping emergency terms. On the other hand, the mapping time stability value reflects the mapping stability of the emergency term mapping model when mapping emergency terms.

[0093] The identification diagnosis value is compared with the identification diagnosis threshold value as follows: If the recognition diagnosis value is less than or equal to the recognition diagnosis threshold, it means that the emergency term mapping model not only has high mapping efficiency but also has relatively stable mapping efficiency when mapping emergency terms, generating stable and efficient signals.

[0094] If the recognition diagnosis value is greater than the recognition diagnosis threshold, it means that the emergency term mapping model not only has low mapping efficiency when mapping emergency terms, but also has relatively unstable mapping efficiency, generating a fluctuating and inefficient signal.

[0095] If a stable and efficient signal is generated, the emergency terminology mapping model will be imported into the data acquisition system, and the emergency knowledge graph will be constructed according to the emergency terminology mapping model.

[0096] If a fluctuating and inefficient signal is generated, the emergency terminology mapping model will be adjusted, and the adjusted emergency terminology mapping model will be re-identified and diagnosed until a stable and efficient signal is generated, imported into the data acquisition system, and the emergency knowledge graph will be constructed according to the emergency terminology mapping model.

[0097] The specific scheme of this embodiment is: within the current test cycle, the emergency terminology mapping model is subjected to emergency terminology recognition diagnosis, the recognition ability of the emergency terminology mapping model is evaluated, and the recognition diagnosis value is obtained, thereby reflecting the recognition ability of the emergency terminology mapping model in the timeliness dimension through the recognition diagnosis value, and then the emergency terminology mapping model is optimized and adjusted in a targeted manner according to the recognition ability result, so as to ensure that the emergency terminology mapping model can perform mapping work in a timely and stable manner in the emergency construction scenario, thereby improving the accuracy of the orthopedic clinical knowledge graph after construction, providing timely decision support for doctors in the emergency construction scenario, and improving the treatment effect of emergency clinical medicine.

[0098] Example 4 See also Figure 2 As shown, the knowledge graph construction system based on orthopedic clinical practice according to the embodiment of the present invention includes the following modules: Feature recognition module: In the previous cycle, the orthopedic data chain in the data repository was extracted and analyzed to obtain the emergency feature range.

[0099] Scenario determination module: During the current test cycle, the orthopedic data chain currently submitted to the data repository is analyzed in the time dimension and compared with the emergency feature range to determine whether the current construction scenario is an emergency construction scenario.

[0100] Model building module: If the scenario is built for emergency care, the emergency terminology in the emergency data in previous cycles and the corresponding professional terminology in the data repository are combined to build an emergency term mapping model.

[0101] Graph construction module: During the current testing cycle, the emergency terminology mapping model is used to perform emergency terminology recognition diagnosis, evaluate the recognition capability of the emergency terminology mapping model, and construct an orthopedic clinical knowledge graph in the emergency construction scenario.

[0102] The basic principles, main features, and advantages of the present invention are shown and described above. Those skilled in the art should understand that the present invention is not limited to the foregoing embodiments. The foregoing embodiments and descriptions are merely illustrative of the principles of the present invention. Various changes and modifications may be made to the present invention without departing from the spirit and scope of the present invention. Such changes and modifications are intended to fall within the scope of the present invention. The scope of protection claimed in the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for constructing a knowledge graph based on orthopedic clinical practice, characterized by: include: In previous cycles, the orthopedic data chain in the data repository was extracted and analyzed to obtain the range of emergency characteristics; During the current test cycle, the orthopedic data chain currently submitted to the data repository is analyzed in terms of time dimension and compared with the emergency feature range to determine whether the current construction scenario is an emergency construction scenario; If building a scenario for emergency care, the emergency terminology in the emergency data of previous cycles is combined with the corresponding professional terminology in the data repository to build an emergency term mapping model; During the current testing cycle, the emergency terminology mapping model is subjected to emergency term recognition diagnosis, the recognition capability of the emergency terminology mapping model is evaluated, and an orthopedic clinical knowledge graph is constructed in the emergency construction scenario.

2. The method for constructing a knowledge graph based on orthopedic clinical practice according to claim 1, characterized in that: The orthopedic data chain in the data repository was extracted and analyzed. The process was as follows: Randomly select an orthopedic data link in the data repository, obtain the entry time of adjacent orthopedic data in the orthopedic data link, perform difference processing, and take the absolute value to output the adjacent entry time interval. If the adjacent entry time interval is less than or equal to the preset adjacent entry time interval, it is an emergency orthopedic data link.

3. The method for constructing a knowledge graph based on orthopedic clinical practice according to claim 1, characterized in that: The process of obtaining the emergency feature range is as follows: Obtain the process submission time of each emergency orthopedic data link, compare them, and extract the maximum and minimum process submission time of the emergency orthopedic data link respectively; Compare the adjacent entry time intervals corresponding to each emergency orthopedic data link, extract the maximum adjacent entry time interval corresponding to each emergency orthopedic data link, and compare all the maximum adjacent entry time intervals to obtain the relative minimum adjacent entry time interval and the relative maximum adjacent entry time interval; The maximum value of the process submission time and the relative maximum adjacent entry time interval are summed, as well as the minimum value of the process submission time and the relative minimum adjacent entry time interval are summed to obtain the maximum value and the minimum value of the emergency feature, and the emergency feature range is obtained.

4. The method for constructing a knowledge graph based on orthopedic clinical practice according to claim 1, characterized in that: The determination process of the emergency construction scenario is as follows: Divide the current test cycle into several test periods, obtain the ratio of the duration of the entire orthopedic data link submission process in each test period to the duration corresponding to the test period, and obtain the process submission duration ratio; In each test period, the entry time of adjacent orthopedic data in the orthopedic data chain is extracted, the difference is processed, and the absolute value is taken to output the current adjacent entry time interval; All the current adjacent input time intervals in the orthopedic data chain are averaged and the average input time interval of the current process is output; The process submission time ratio is summed with the average of the current process entry time intervals to obtain the scenario evaluation value. If the scenario evaluation value is within the emergency feature range, the corresponding construction scenario is marked as an emergency construction scenario.

5. The method for constructing a knowledge graph based on orthopedic clinical practice according to claim 1, characterized in that: Divide the previous period into several previous periods. Within the previous period, randomly extract an emergency term from the emergency orthopedics data chain as the target term. Obtain the professional term corresponding to the target term in the data repository within the previous period, and combine them according to the time series to obtain multiple groups of term correspondences. Among multiple terminology correspondence groups, extract the terminology correspondence groups that have the same professional terminology, and obtain the proportion of the total number of all terminology correspondence groups to obtain the corresponding number ratio of the same professional terminology.

6. The method for constructing a knowledge graph based on orthopedic clinical practice according to claim 5, characterized in that: Based on the terminology corresponding groups with the same professional term, all previous time periods in which the corresponding term appears are obtained as the same professional term period, and are sorted according to the time series to obtain multiple same professional term period sequences. The sorting of all adjacent same professional term periods in each same professional term period sequence in the previous cycle is input into the convolution layer of the convolutional neural network to obtain multiple training convolution layers; The order of adjacent periods of the same professional terminology in multiple training convolutional layers in previous cycles is input into the improved Euclidean distance calculation formula, and the output is the corresponding stable value of the term.

7. The method for constructing a knowledge graph based on orthopedic clinical practice according to claim 1, characterized in that: The construction process of the emergency terminology mapping model is as follows: Perform geometric product calculation on the quantity ratio of the same professional term and the corresponding stable value of the term, and output the corresponding term mapping value; If the corresponding term mapping value is greater than or equal to the corresponding term mapping threshold, a mapping stability signal is generated, and an emergency term mapping model is constructed based on multiple training convolutional layers corresponding to the generated mapping stability signals.

8. The method for constructing a knowledge graph based on orthopedic clinical practice according to claim 1, characterized in that: In any test period, the mapping duration between any test emergency term and any test professional term is obtained, the proportion of the total test period duration is calculated, and the average processing is performed to output the period mapping duration ratio; The time period mapping duration ratios corresponding to all test periods in the current test cycle are averaged and the mapping duration test value is output; The mapping duration ratios within each test period are sorted according to the corresponding test time to construct a period test sequence. The period test sequences corresponding to adjacent test periods are input into the Pearson distance formula, and then averaged and calculated to obtain the stable value of the mapping duration.

9. The method for constructing a knowledge graph based on orthopedic clinical practice according to claim 8, characterized in that: If the recognition diagnosis value is less than or equal to the recognition diagnosis threshold, a stable and efficient signal is generated, the emergency term mapping model is imported into the data acquisition system, and the emergency knowledge graph is constructed according to the emergency term mapping model; If the recognition diagnosis value is greater than the recognition diagnosis threshold, a fluctuating inefficient signal is generated, the emergency term mapping model is adjusted, and the adjusted emergency term mapping model is re-identified and diagnosed until a stable and efficient signal is generated, imported into the data acquisition system, and the emergency knowledge graph is constructed according to the emergency term mapping model.

10. A knowledge graph construction system based on orthopedic clinical practice, characterized by: include: Feature recognition module: In the previous cycle, the orthopedic data chain in the data repository was extracted and analyzed to obtain the emergency feature range; Scenario determination module: During the current test cycle, the orthopedic data chain currently submitted to the data repository is analyzed in the time dimension, and compared with the emergency feature range to determine whether the current construction scenario is an emergency construction scenario; Model building module: If the scenario is built for emergency care, the emergency terminology in the emergency data of previous cycles is combined with the corresponding professional terminology in the data repository to build an emergency term mapping model; Graph construction module: During the current testing cycle, the emergency terminology mapping model is used to perform emergency terminology recognition diagnosis, evaluate the recognition capability of the emergency terminology mapping model, and construct an orthopedic clinical knowledge graph in the emergency construction scenario.

Images