Medical decision auxiliary method and system based on medical rules

Abstract

The invention discloses a medical decision assistance method and system based on medical rules, and relates to the technical field of medical decision assistance, and the method comprises the following steps: receiving medical data of a patient from various medical devices and information systems; standardizing the received data, and mapping the standardized data to a corresponding medical concept after the data is standardized; and in the data mapping process, acquiring and recording associated information of the data. According to the invention, through standardized processing and machine learning model analysis, the problem of misassociation during association of medical data from different sources is effectively solved, and the accuracy of diagnosis and treatment is improved. And misdiagnosis and excessive treatment are prevented by evaluating the data association quality and screening out misassociated data. And abnormal data identification and elimination are optimized by using a clustering algorithm, so that diagnosis is ensured to be based on reliable data, unnecessary medical intervention caused by data errors is reduced, and the overall nursing quality and the medical service level are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of medical decision assistance, and in particular to a medical decision assistance method and system based on medical rules. Background Art

[0002] The medical decision support system based on medical rules is an intelligent decision support tool that receives patient medical data from different devices and medical systems, standardizes the data and maps it to medical concepts related to auxiliary diagnosis. Through the built-in rule engine, the system evaluates these medical concept data and triggers the corresponding medical rules, thereby giving prompts, warnings and specific treatment suggestions on the interface to help doctors make more accurate and timely decisions. At the same time, the system records the doctor's handling of these prompts and suggestions for subsequent performance improvement and personalized optimization, thereby improving the system's decision support capabilities.

[0003] Mapping data to medical concepts means matching standardized data with predefined medical concepts or diagnostic frameworks. Medical concepts are a set of standardized terms and definitions used to describe specific pathologies, physiological states, symptoms, signs, diagnoses, and treatments. For example, SNOMED CT and ICD-10 are widely used medical terminology systems that unify the representation of medical concepts.

[0004] When data is standardized and mapped to medical concepts related to auxiliary diagnosis, data from different sources need to be linked together to form a complete medical concept. If these data are not correctly linked, inaccurate mapping may result. For example, incorrectly linking blood pressure and heart rate data at different time points may cause doctors to misjudge the patient's true health status. The patient may be in a state of hypertension at one point in time, and the heart rate suddenly increases at another time, but these changes occur in different scenarios. If the data is mistakenly thought to be recorded at the same time, the doctor may draw the wrong conclusion that the patient is at risk of acute cardiovascular events, such as myocardial infarction or stroke, leading to unnecessary emergency intervention or overtreatment. On the contrary, if the doctor fails to notice that the data comes from different time points, potential health risks may be ignored, resulting in delayed diagnosis.

[0005] The above information disclosed in this Background section is only for enhancement of understanding of the background of the present disclosure and therefore it may contain information that does not constitute the prior art that is already known to one of ordinary skill in the art. Summary of the invention

[0006] The purpose of the present invention is to provide a medical decision-making assistance method and system based on medical rules. Through standardized processing and intelligent machine learning model analysis, it effectively solves the problem of misassociation that may occur in the association process of medical data from different sources, improves the accuracy of diagnosis and treatment, and prevents the risk of misdiagnosis, overtreatment or delayed diagnosis by accurately evaluating the quality of data association and timely screening out misassociated data. At the same time, the clustering algorithm is used to further optimize the identification and exclusion of abnormal data, ensuring that diagnosis and treatment decisions are based on accurate and reliable data, which not only improves the accuracy and effectiveness of medical decisions, but also reduces unnecessary medical interventions caused by data association errors, and overall improves the quality of patient care and the level of medical services to solve the problems in the above-mentioned background technology.

[0007] In order to achieve the above object, the present invention provides the following technical solution: a medical decision-making assistance method based on medical rules, comprising the following steps:

[0008] Receive patient medical data from various medical devices and information systems;

[0009] Standardize the received data, and after the data is standardized, map the standardized data to corresponding medical concepts;

[0010] In the process of data mapping, obtain and record the associated information of the data;

[0011] After obtaining the associated information, the associated information is analyzed to preliminarily identify potential mis-associated data, and the identified data is screened out to provide accurate input data for the prediction model;

[0012] Use pre-trained machine learning models to intelligently predict the association of data from different sources, and further determine whether there are anomalies in data association based on the prediction results of the machine learning model;

[0013] When anomalies in data association are detected, a clustering algorithm is used to conduct a comprehensive analysis of data from different sources. Based on the results of the clustering algorithm, the data with abnormal associations are identified, marked, and excluded from the final diagnosis and treatment decisions.

[0014] Analyze the acquired association information, preliminarily identify potential mis-associated data, and filter out the identified data. The specific steps are as follows:

[0015] Perform a timestamp and source device consistency check on each piece of data to ensure that all associated data points are collected within the correct time range and come from the same or compatible devices;

[0016] Conduct context and situation analysis of data to confirm whether the data collection conditions and patient status are consistent, and ensure that all related data are acquired under similar or identical collection conditions and patient status;

[0017] Test the rationality and medical logic of the data, check whether there is a relationship between the data that conforms to medical common sense, and ensure that the relationship and pattern between the data conform to known medical common sense and physiological laws;

[0018] Screen and clean the potentially mis-associated data that are initially identified.

[0019] Preferably, a pre-trained machine learning model is used to make real-time predictions on the association of data from different sources. The specific steps are:

[0020] The key parameters for associating data from different sources are obtained, including the time synchronization information of data from different devices and the difference in data sampling frequencies of different devices. After analyzing the time synchronization information of data from different devices and the difference in data sampling frequencies of different devices, a time synchronization index and a sampling frequency difference index are generated respectively. The time synchronization index and the sampling frequency difference index are predicted using a pre-trained machine learning model, and an association evaluation coefficient is generated based on the model. The association of data from different sources is intelligently predicted through the association evaluation coefficient.

[0021] Preferably, the correlation evaluation coefficient generated by intelligently predicting the correlation of data from different sources through a machine learning model is compared and analyzed with a preset correlation evaluation coefficient reference threshold to further determine whether there is an abnormality in data correlation. The specific steps are as follows:

[0022] If the correlation evaluation coefficient is less than the correlation evaluation coefficient reference threshold, a correlation hidden danger signal is generated, indicating that data from different sources have the hidden danger of abnormal correlation;

[0023] If the correlation evaluation coefficient is greater than or equal to the correlation evaluation coefficient reference threshold, an efficient correlation signal is generated, indicating that data from different sources can be efficiently correlated.

[0024] After analyzing the time synchronization information of data from different devices, the specific steps for generating the time synchronization index are as follows:

[0025] For each pair of associated data points from different devices, calculate the absolute difference in their timestamps. The timestamp of the data from device A is T A , the data timestamp of device B is T A , then the timestamp difference calculation formula is: D = |T A -T B|, where D represents the timestamp difference, reflecting the time interval between the two data points;

[0026] The time difference weight factor is introduced to further quantify the impact of the time difference. The time difference weight factor is inversely proportional to the timestamp difference, and the calculation expression is: W D is the time difference weight factor, ranging from 0 to 0.5, which is used to measure the impact of timestamp differences. α is the first adjustment parameter, which controls the sensitivity of the weight factor to timestamp differences.

[0027] Calculate the device weight factor to reflect the synchronization quality between devices. The device weight factor is calculated based on the known delay or transmission time of the device. The calculation expression is: W E is the device weight factor, ranging from 0 to 0.5, indicating the synchronization quality between devices, β is the second adjustment parameter, controlling the sensitivity of the device weight factor to latency, and L is the known latency or data transmission duration between devices;

[0028] Comprehensive time difference weight factor W D and the device weight factor W E , calculate the time synchronization index, which reflects the overall synchronization degree of the data in time. The calculation expression is: TSI index =W D ·W E , TSI index It is the time synchronization index, which indicates the degree of time synchronization.

[0029] After analyzing the differences in data sampling frequencies of different devices, the specific steps for generating the sampling frequency difference index are as follows:

[0030] Collect data sampling frequencies from different devices. Device A and device B are used to measure the patient’s physiological parameters, respectively. A and f B Respectively represent the sampling frequencies of device A and device B;

[0031] Calculate the sampling frequency ratio between the two devices. The frequency ratio is calculated by the following formula: In the formula, R AB Indicates the sampling frequency ratio of device A and device B, max(f A , f B ) represents the highest sampling frequency of device A and device B, min(f A , f B ) represents the lowest sampling frequency of device A and device B;

[0032] Calculate the absolute value of the sampling frequency difference between device A and device B to further understand the degree of frequency difference. The calculation expression for the absolute difference in sampling frequency is: AB =|f A -f B |, D AB Indicates the absolute difference in sampling frequency between device A and device B;

[0033] The sampling frequency ratio R of device A and device B AB The absolute difference D between the sampling frequencies of device A and device B AB Calculate the sampling frequency difference index. The expression is: Where SFDI index is the sampling frequency difference index, which indicates the degree of sampling frequency difference between devices, ln(1+R AB ) uses a logarithmic function to smooth the frequency ratio, The absolute value of the frequency difference is adjusted using a square root function to enhance the perception of the magnitude of the difference.

[0034] A clustering algorithm is used to conduct a comprehensive analysis of data from different sources. Based on the results of the clustering algorithm, data with abnormal correlations is identified. The specific steps are as follows:

[0035] When anomalies in data association are detected, all relevant data points and their corresponding time synchronization index and sampling frequency difference index are collated;

[0036] After the data is prepared, select a clustering algorithm to analyze the data;

[0037] When performing cluster analysis, the data are grouped according to the input time synchronization index, sampling frequency difference index, and the standardized medical concept data embedding generated by mapping, thereby forming different cluster groups;

[0038] Based on the results of the clustering algorithm, identify data points with abnormal associations;

[0039] Data marked as abnormal were excluded from the final diagnosis and treatment decision.

[0040] Preferably, based on the results of the clustering algorithm, data points with abnormal associations are identified, and the specific steps are as follows:

[0041] Calculate the closeness within each group to assess the similarity between data points;

[0042] The separation between groups was calculated to assess the differences between different groups;

[0043] By comparing the internal compactness of the group and the separation between the groups, abnormal data points are identified. First, an upper threshold of the internal compactness and a lower threshold of the separation between the groups are set. For the internal compactness of the group, if the internal compactness of a data point or group exceeds the preset upper threshold, the data point is marked as an abnormal point; at the same time, if the separation between the data point or group and other groups is lower than the preset lower threshold, the data point is also marked as an abnormal point.

[0044] A medical decision support system based on medical rules, including a data receiving module, a data standardization and mapping module, a related information acquisition module, a related information analysis and screening module, an intelligent prediction and anomaly detection module, and an anomaly analysis and processing module;

[0045] A data receiving module receives the patient's medical data from various medical devices and information systems;

[0046] The data standardization and mapping module standardizes the received data and maps the standardized data to the corresponding medical concepts after data standardization;

[0047] The associated information acquisition module acquires and records the associated information of the data during the data mapping process;

[0048] The correlation information analysis and screening module analyzes the obtained correlation information after obtaining it, preliminarily identifies potential mis-correlated data, and screens out the identified data to provide accurate input data for the prediction model;

[0049] The intelligent prediction and anomaly detection module uses a pre-trained machine learning model to intelligently predict the association of data from different sources, and further determines whether there are anomalies in the data association based on the prediction results of the machine learning model;

[0050] The abnormal analysis and processing module uses a clustering algorithm to conduct a comprehensive analysis of data from different sources when an abnormality is detected in data association. Based on the results of the clustering algorithm, the data with abnormal associations is identified, marked, and excluded from the final diagnosis and treatment decision.

[0051] In the above technical solution, the technical effects and advantages provided by the present invention are:

[0052] The medical decision-making assistance method based on medical rules of the present invention effectively solves the problem of mis-association that may occur in the association process of medical data from different sources, thereby significantly improving the accuracy of diagnosis and treatment. Through data standardization processing and intelligent machine learning model analysis, it is possible to accurately evaluate the association quality between data from different sources, timely identify and screen out potential mis-associated data, avoid erroneously associating data collected at different time points or in different situations, and thus prevent the risk of misdiagnosis, overtreatment or delayed diagnosis caused by this.

[0053] The present invention further optimizes the process of identifying and eliminating abnormal data through clustering algorithms, ensuring that the final diagnosis and treatment decisions are based only on accurate and reliable data. This not only improves the accuracy and effectiveness of medical decisions, but also reduces the risk of patients receiving unnecessary medical interventions due to data association errors, ultimately improving the quality of patient care and the overall level of medical services. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0055] Figure 1 This is a method flow chart of the medical decision-making assistance method based on medical rules of the present invention.

[0056] Figure 2 It is a module schematic diagram of the medical decision support system based on medical rules of the present invention. DETAILED DESCRIPTION

[0057] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as limited to the examples set forth herein; rather, these example embodiments are provided so that the description of the present disclosure will be more comprehensive and complete, and the concept of the example embodiments will be fully conveyed to those skilled in the art.

[0058] The present invention provides Figure 1 The medical decision-making assistance method based on medical rules shown includes the following steps:

[0059] Receive patient medical data from various medical devices and information systems;

[0060] These data may include blood pressure, heart rate, blood sugar levels, laboratory test results, etc., and the formats and units may vary. The purpose of standardization is to convert these data from different sources into a unified format and unit to ensure the compatibility and comparability of the data. Standardization may include converting data units (for example, converting blood sugar from mg / dL to mmol / L), time synchronization (aligning records from different sources to a unified timeline), and data cleaning (removing erroneous, missing, or duplicate data).

[0061] Standardize the received data, and after the data is standardized, map the standardized data to corresponding medical concepts;

[0062] For example, blood pressure readings might be mapped to a status of "hypertension" or "normotension," and heart rate data might be mapped to a status of "normal heart rate" or "tachycardia." This step uses predefined medical standards and rules, such as SNOMED CT or ICD-10, to ensure that all data is accurately matched to standardized medical concepts. This process helps the system understand the medical significance of the data, preparing it for further analysis and decision support.

[0063] In the process of data mapping, obtain and record the associated information of the data;

[0064] Related information includes the source of the data, the time of acquisition, the type of device that generated the data, and the collection environment. The acquisition of related information is critical because it helps to identify and distinguish data from different time points or different sources. If this information is missing or inaccurate, the system may mistakenly combine unrelated data, leading to incorrect medical judgments.

[0065] After obtaining the associated information, the associated information is analyzed to preliminarily identify potential mis-associated data, and the identified data is screened out to provide accurate input data for the prediction model;

[0066] Analyze the acquired association information, preliminarily identify potential mis-associated data, and filter out the identified data. The specific steps are as follows:

[0067] Perform a timestamp and source device consistency check on each piece of data to ensure that all associated data points are collected within the correct time range and come from the same or compatible devices;

[0068] The consistency check of the timestamp and source device of each piece of data is intended to ensure that the data collection process is reliable and consistent. Specifically, this check includes verifying the timestamp of each piece of data to confirm whether the data is collected within a reasonable time frame to avoid mistaking data from different time periods for events that occurred at the same time. In addition, it is necessary to check the source device of the data to ensure that the data comes from the same or functionally compatible devices to prevent errors or incompatibilities caused by device differences. If the data comes from incompatible devices, there may be association errors in the data, affecting the authenticity of the data and the accuracy of the analysis. Therefore, this check is a key step to ensure data quality and the reliability of subsequent analysis.

[0069] Conduct context and situation analysis of data to confirm whether the data collection conditions and patient status are consistent, and ensure that all related data are acquired under similar or identical collection conditions and patient status;

[0070] For example, if one set of data is collected when the patient is resting, and another set of data is collected when the patient is moving, then even if the timestamps of these data are close, the direct association between them may still be inaccurate. By analyzing the contextual information of the data (such as the patient's activity status, environmental conditions, equipment settings, etc. at the time of collection), the system can identify those mis-associated data that are close in time but actually have no correlation, thereby further screening out potential abnormal data points.

[0071] Test the rationality and medical logic of the data, check whether there is a relationship between the data that conforms to medical common sense, and ensure that the relationship and pattern between the data conform to known medical common sense and physiological laws;

[0072] This step can avoid incorrect diagnosis and treatment decisions caused by abnormal or illogical data association. For example, if the heart rate data shows that the patient is in an extremely high heart rate state, but the blood pressure data shows that it is completely normal, this combination may not conform to medical logic, indicating that there may be data association errors. The system automatically detects contradictions and anomalies between data by setting reasonable medical thresholds and logical rules, thereby preliminarily identifying incorrectly associated data that does not conform to medical logic.

[0073] Screen and clean up the potentially mis-associated data that are initially identified;

[0074] The purpose of screening is to remove data points that have been marked as abnormal or unreasonable to ensure that the data remaining in the system is as accurate and reliable as possible. These screened data may be moved to a separate storage area for subsequent manual review or further analysis. At the same time, the system will also record the characteristics and causes of these misassociated data to help improve future data processing algorithms and rules, thereby further optimizing the quality of input data for predictive models. This cleaning process ensures the accuracy and consistency of the input data of subsequent machine learning models, laying the foundation for accurate prediction and decision-making.

[0075] Use pre-trained machine learning models to intelligently predict the association of data from different sources, and further determine whether there are anomalies in data association based on the prediction results of the machine learning model;

[0076] Use pre-trained machine learning models to make real-time predictions on data associations from different sources. The specific steps are:

[0077] The key parameters for associating data from different sources are obtained, including the time synchronization information of data from different devices and the difference in data sampling frequencies of different devices. After analyzing the time synchronization information of data from different devices and the difference in data sampling frequencies of different devices, a time synchronization index and a sampling frequency difference index are generated respectively. The time synchronization index and the sampling frequency difference index are predicted using a pre-trained machine learning model, and an association evaluation coefficient is generated based on the model. The association of data from different sources is intelligently predicted through the association evaluation coefficient.

[0078] A pre-trained machine learning model refers to a model that has been trained and optimized using a large amount of historical data before actual application. This model learns how to process and analyze specific types of data by learning patterns and rules in historical data. In the present invention, the pre-trained machine learning model is trained by a large number of historical data sets with known association quality, and the model learns how the differences in time synchronization and sampling frequency of the data affect the reliability and accuracy of data association. The training process includes multiple iterations, and its performance in the data association prediction task is gradually improved by adjusting the parameters of the model. When the training reaches a certain level, the model can effectively predict the quality of data association intelligently based on the input parameters such as the time synchronization index and the sampling frequency difference index, and generate an association evaluation coefficient. This pre-trained model does not need to be retrained when applied to new data, and can immediately provide effective prediction results, greatly improving the efficiency and accuracy of the system in real-time data processing. Therefore, the use of a pre-trained model can quickly and accurately analyze new data association situations and help the system identify potential anomalies or errors.

[0079] When data from different devices are poorly synchronized in time, this indicates that there are potential anomalies in the association of data from different sources, because low time synchronization may mean that the data were not collected in the same time period or during the same event. Data collected at different times may reflect different physiological states or environmental conditions of patients. Incorrectly associating data from these different time points may lead to incorrect diagnosis and treatment decisions, thereby affecting patient safety and medical outcomes. Therefore, ensuring the time synchronization of data is crucial for accurate medical data association.

[0080] After analyzing the time synchronization information of data from different devices, the specific steps for generating the time synchronization index are as follows:

[0081] For each pair of associated data points from different devices, calculate the absolute difference in their timestamps. The timestamp of the data from device A is T A , the data timestamp of device B is T A , then the timestamp difference calculation formula is: D = |T A -T B |, where D represents the timestamp difference, reflecting the time interval between the two data points;

[0082] A larger timestamp difference indicates a lower degree of time synchronization, and a smaller timestamp difference indicates a higher degree of time synchronization.

[0083] In order to further quantify the impact of time difference, a time difference weight factor is introduced. The time difference weight factor is inversely proportional to the timestamp difference. The calculation expression is: W D is the time difference weight factor, ranging from 0 to 0.5, which is used to measure the impact of timestamp differences. α is the first adjustment parameter, which controls the sensitivity of the weight factor to timestamp differences.

[0084] A larger first adjustment parameter makes the weight factor more sensitive to time differences.

[0085] Data synchronization between different devices may be affected by factors such as device performance and data transmission delay. Therefore, a device weight factor needs to be calculated to reflect the synchronization quality between devices. The device weight factor is calculated based on the known delay or transmission time of the device. The calculation expression is: W E is the device weight factor, ranging from 0 to 0.5, indicating the synchronization quality between devices. β is the second adjustment parameter, which controls the sensitivity of the device weight factor to latency. L is the known latency or data transmission duration between devices, which can be obtained from device specifications or network transmission characteristics.

[0086] Comprehensive time difference weight factor W Dand the device weight factor W E , calculate the time synchronization index, which reflects the overall synchronization degree of the data in time. The calculation expression is: TSI index =W D ·W E , TSI index is the time synchronization index, which indicates the degree of time synchronization;

[0087] The closer the time synchronization index is to 0.25, the better the time synchronization is; the closer it is to 0, the worse the time synchronization is.

[0088] It can be seen from the time synchronization index that the smaller the performance value of the time synchronization index generated after analyzing the time synchronization information of data from different devices, the worse the degree of time synchronization of data from different devices, which in turn indicates that there is a greater risk of potential anomalies when these data are associated. This is because a lower time synchronization index indicates a large difference between data collection time points or poor synchronization quality between devices, which may cause the data to be not in the same time frame and thus be incorrectly associated together. On the contrary, a higher time synchronization index indicates that the data is better synchronized in time, the potential anomaly risk is smaller when associated, and the data is more correlated and consistent, which helps to more accurately diagnose and support decisions.

[0089] When there are significant differences in the data sampling frequencies of different devices, this indicates that there may be potential anomalies when associating data from different sources, because different sampling frequencies will lead to inconsistencies in the time dimension of the data, thus affecting the synchronization and comparability of the data. If these data are indiscriminately associated, it may confuse the trend of data changes at different time points, leading to incorrect medical analysis and judgment. For example, heart rate data is sampled at a frequency of once per second, while blood pressure data is sampled once per minute. This frequency difference may cause the data to not accurately reflect the patient's true physiological state when associated, increasing the risk of misdiagnosis and incorrect treatment.

[0090] After analyzing the differences in data sampling frequencies of different devices, the specific steps for generating the sampling frequency difference index are as follows:

[0091] Collect data sampling frequencies from different devices. Device A and device B are used to measure the patient’s physiological parameters, such as heart rate and blood pressure, respectively. A and f B Respectively represent the sampling frequency of device A and device B, in Hz (sampling times per second);

[0092] Calculate the sampling frequency ratio between the two devices. The sampling frequency ratio is to standardize the frequency differences between different devices so that they can be compared on the same scale. The frequency ratio is calculated using the following formula: In the formula, RAB Indicates the sampling frequency ratio of device A and device B, max(f A , f B ) represents the highest sampling frequency of device A and device B, min(f A , f B ) represents the lowest sampling frequency of device A and device B;

[0093] Calculating the absolute value of the sampling frequency difference between device A and device B can help further understand the degree of frequency difference. The calculation expression for the absolute difference in sampling frequency is: AB =|f A -f B |, D AB Indicates the absolute difference in sampling frequency between device A and device B;

[0094] The sampling frequency ratio R of device A and device B AB The absolute difference D between the sampling frequencies of device A and device B AB Calculate the sampling frequency difference index. The expression is: Where SFDI index is the sampling frequency difference index, which indicates the degree of sampling frequency difference between devices, ln(1+R AB ) Use a logarithmic function to smooth the frequency ratio and reduce the impact of extreme ratios. The absolute value of the frequency difference is adjusted using a square root function to enhance the perception of the magnitude of the difference.

[0095] From the sampling frequency difference index, we can see that the larger the performance value of the sampling frequency difference index generated after analyzing the differences in data sampling frequencies of different devices, the more significant the differences in data sampling frequencies between different devices are, which means that there is a greater potential for abnormalities when associating these data together, because different sampling frequencies may cause inconsistencies in the time dimension of the data, thereby affecting the synchronization and accuracy of the data. Therefore, a larger sampling frequency difference index indicates that the association of these data needs to be handled with caution to avoid misjudgment or erroneous analysis due to frequency differences. Conversely, a smaller sampling frequency difference index indicates that the data sampling frequencies are closer, the potential for abnormalities when associating data is smaller, and the synchronization and consistency of the data are more guaranteed.

[0096] The machine learning model is not specifically limited here, and can achieve the time synchronization index TSI index and sampling frequency difference index SFDI index Perform comprehensive analysis to generate association evaluation coefficients τ In order to realize the technical solution of the present invention, the present invention provides a specific implementation method;

[0097] Association evaluation coefficient τ The resulting calculation formula is:

[0098]

[0099] , where d1 and d2 are time synchronization index TSI respectively. index and sampling frequency difference index SFDI index The preset proportional coefficient, and d1 and d2 are both greater than 0.

[0100] It can be seen from the correlation evaluation coefficient that the larger the performance value of the time synchronization index generated after analyzing the time synchronization information of data from different devices, the smaller the performance value of the sampling frequency difference index generated after analyzing the difference in data sampling frequency of different devices. In other words, the larger the performance value of the correlation evaluation coefficient generated by intelligently predicting the correlation of data from different sources through a machine learning model, the smaller the risk of potential anomalies when associating these data together, and vice versa.

[0101] The correlation evaluation coefficient generated by intelligently predicting the correlation of data from different sources through the machine learning model is compared and analyzed with the pre-set correlation evaluation coefficient reference threshold to further determine whether there is anomaly in data correlation. The specific steps are as follows:

[0102] If the correlation evaluation coefficient is less than the correlation evaluation coefficient reference threshold, a correlation hidden danger signal is generated, indicating that data from different sources have the hidden danger of abnormal correlation;

[0103] If the correlation evaluation coefficient is greater than or equal to the correlation evaluation coefficient reference threshold, an efficient correlation signal is generated, indicating that data from different sources can be efficiently correlated.

[0104] When abnormalities in data association are detected, a clustering algorithm is used to conduct a comprehensive analysis of data from different sources. Based on the results of the clustering algorithm, data with abnormal associations are identified, marked, and excluded from the final diagnosis and treatment decision;

[0105] A clustering algorithm is used to conduct a comprehensive analysis of data from different sources. Based on the results of the clustering algorithm, data with abnormal correlations is identified. The specific steps are as follows:

[0106] When anomalies in data association are detected, all relevant data points and their corresponding time synchronization index, sampling frequency difference index, and mapping are used to obtain standardized medical concept data embedding.

[0107] The time synchronization index is used to quantify the degree of temporal synchronization of data from different sources, while the sampling frequency difference index evaluates the differences in sampling frequencies between devices; the medical concept is used to evaluate the differences between different symptoms and treatments.

[0108] Medical concept data embedding is obtained by pre-training wordvec on medical knowledge data, and the cluster embedding is the mean of all medical concept data embeddings corresponding to the data point.

[0109] These characteristic values ​​provide basic data for cluster analysis, enabling the clustering algorithm to effectively group data based on these indicators and distinguish normal data from abnormal data.

[0110] After the data is prepared, select a clustering algorithm to analyze the data;

[0111] Common clustering algorithms include K-means clustering, density clustering (DBSCAN), hierarchical clustering, etc. According to the characteristics of the data and the focus of the analysis, the most appropriate algorithm can be selected. For example, if the distribution of the data is complex and has an irregular shape, density clustering (DBSCAN) is more appropriate because it does not depend on the shape and quantity of the data. Choosing a suitable clustering algorithm can improve the accuracy of abnormal data detection and make it easier for abnormal data to be correctly marked.

[0112] When performing cluster analysis, the data are grouped according to the input time synchronization index, sampling frequency difference index, and the standardized medical concept data embedding generated by mapping, thereby forming different cluster groups;

[0113] In these clusters, the algorithm identifies those outliers that are clearly different from other data points. Outliers may be characterized by significant differences in time synchronization, large differences in sampling frequency, etc. Through cluster analysis, the system can discover potential patterns and outlier groups in the data, which often represent potential problems or inconsistencies in the data association process.

[0114] Based on the results of the clustering algorithm, identify data points with abnormal associations;

[0115] Among them, the clustering algorithm groups most normal data together, while abnormal data will form independent isolated points or abnormal groups. According to the abnormal group, the data points in it are marked as abnormal data. This step ensures that the abnormal data will not be misused in the subsequent diagnosis and treatment decision-making process, thereby avoiding incorrect assessment and decision-making of the patient's health status. These marked data will also be recorded for further analysis and improvement of the data processing process.

[0116] Exclude data marked as abnormal from final diagnosis and treatment decisions;

[0117] The purpose of excluding abnormal data is to ensure that all data used for diagnosis and treatment decisions are reliable and accurate, and to avoid misdiagnosis or overtreatment due to incorrect data association. At the same time, the system can also conduct feedback analysis based on these excluded abnormal data to optimize future data association and processing algorithms. This continuous improvement mechanism helps to improve the overall reliability and decision-making quality of the system and ensure that patients receive the best medical services.

[0118] Based on the results of the clustering algorithm, identify the data points with abnormal associations. The specific steps are as follows:

[0119] Calculate the closeness within each group to assess the similarity between data points;

[0120] The intra-group closeness is usually measured by calculating the distance between the data points within the group. A low intra-group closeness indicates that the data points within the group are similar to each other, indicating that these points belong to the same class or have similar characteristics. If some data points in a group are significantly farther away from other data points, these data points may be outliers. The distance between data points can be calculated using methods such as Euclidean distance or Manhattan distance.

[0121] The separation between groups was calculated to assess the differences between different groups;

[0122] Separation is measured by calculating the distance between groups, usually expressed as the distance between the group centroids (i.e., the average position of the data points within the group). If the separation between two groups is small, it indicates that their data points are close in feature space and may represent the same type of data or groups with very similar features. For anomaly detection, if a data point or group is significantly far away from other groups, it means it is an abnormal group or outlier.

[0123] By comparing the internal compactness of the group and the separation between the groups, the abnormal data points are identified. First, an upper threshold of the internal compactness and a lower threshold of the separation between the groups are set. For the internal compactness of the group, if the internal compactness of a data point or a group exceeds the preset upper threshold, it indicates that the distance between the data point and other data points in the group is too large, and the data point is marked as an abnormal point; at the same time, if the separation between the data point or group and other groups is lower than the preset lower threshold, it indicates that the difference between the data point and other groups is not significant, and the data point is also marked as an abnormal point;

[0124] Through this threshold comparison method, data points that significantly deviate from the normal pattern can be effectively screened out and marked as abnormal.

[0125] The medical decision-making assistance method based on medical rules of the present invention effectively solves the problem of mis-association that may occur in the association process of medical data from different sources, thereby significantly improving the accuracy of diagnosis and treatment. Through data standardization processing and intelligent machine learning model analysis, it is possible to accurately evaluate the association quality between data from different sources, timely identify and screen out potential mis-associated data, avoid erroneously associating data collected at different time points or in different situations, and thus prevent the risk of misdiagnosis, overtreatment or delayed diagnosis caused by this. In addition, the clustering algorithm further optimizes the identification and exclusion process of abnormal data, ensuring that the final diagnosis and treatment decisions are based only on accurate and reliable data, which not only improves the accuracy and effectiveness of medical decisions, but also reduces the risk of patients receiving unnecessary medical interventions due to data association errors, and ultimately improves the quality of patient care and the overall level of medical services.

[0126] The present invention provides Figure 2 The medical decision support system based on medical rules shown includes a data receiving module, a data standardization and mapping module, a related information acquisition module, a related information analysis and screening module, an intelligent prediction and anomaly detection module, and an anomaly analysis and processing module;

[0127] A data receiving module receives the patient's medical data from various medical devices and information systems;

[0128] The data standardization and mapping module standardizes the received data and maps the standardized data to the corresponding medical concepts after data standardization;

[0129] The associated information acquisition module acquires and records the associated information of the data during the data mapping process;

[0130] The correlation information analysis and screening module analyzes the obtained correlation information after obtaining it, preliminarily identifies potential mis-correlated data, and screens out the identified data to provide accurate input data for the prediction model;

[0131] The intelligent prediction and anomaly detection module uses a pre-trained machine learning model to intelligently predict the association of data from different sources, and further determines whether there are anomalies in the data association based on the prediction results of the machine learning model;

[0132] The abnormal analysis and processing module uses a clustering algorithm to conduct a comprehensive analysis of data from different sources when abnormalities are detected in data associations. Based on the results of the clustering algorithm, the data with abnormal associations are identified, marked, and excluded from the final diagnosis and treatment decisions;

[0133] The medical decision assistance method based on medical rules provided in the embodiment of the present invention is implemented through the above-mentioned medical decision assistance system based on medical rules. The specific methods and processes of the medical decision assistance system based on medical rules are detailed in the embodiment of the above-mentioned medical decision assistance method based on medical rules, which will not be repeated here.

[0134] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0135] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0136] The above description is only by way of illustration of certain exemplary embodiments of the present invention. It is undoubted that those skilled in the art can modify the described embodiments in various ways without departing from the spirit and scope of the present invention. Therefore, the above drawings and descriptions are illustrative in nature and should not be construed as limiting the scope of protection of the claims of the present invention.

Claims

1. A medical decision support method based on medical rules, characterized in that: The following steps are involved: Receive patient medical data from various medical devices and information systems; Standardize the received data, and after the data is standardized, map the standardized data to corresponding medical concepts; In the process of data mapping, obtain and record the associated information of the data; After obtaining the associated information, the associated information is analyzed to preliminarily identify potential mis-associated data, and the identified data is screened out to provide accurate input data for the prediction model; Use pre-trained machine learning models to intelligently predict the association of data from different sources, and further determine whether there are anomalies in data association based on the prediction results of the machine learning model; When anomalies in data association are detected, a clustering algorithm is used to conduct a comprehensive analysis of data from different sources. Based on the results of the clustering algorithm, the data with abnormal associations are identified, marked, and excluded from the final diagnosis and treatment decisions.

2. The medical decision-making assistance method based on medical rules according to claim 1, characterized in that: Analyze the acquired association information, preliminarily identify potential mis-associated data, and filter out the identified data. The specific steps are as follows: Perform a timestamp and source device consistency check on each piece of data to ensure that all associated data points are collected within the correct time range and come from the same or compatible devices; Conduct context and situation analysis of data to confirm whether the data collection conditions and patient status are consistent, and ensure that all related data are acquired under similar or identical collection conditions and patient status; Test the rationality and medical logic of the data, check whether there is a relationship between the data that conforms to medical common sense, and ensure that the relationship and pattern between the data conform to known medical common sense and physiological laws; Screen and clean the potentially mis-associated data that are initially identified.

3. The medical decision-making assistance method based on medical rules according to claim 2, characterized in that: Use pre-trained machine learning models to make real-time predictions on data associations from different sources. The specific steps are: The key parameters for associating data from different sources are obtained, including the time synchronization information of data from different devices and the difference in data sampling frequencies of different devices. After analyzing the time synchronization information of data from different devices and the difference in data sampling frequencies of different devices, a time synchronization index and a sampling frequency difference index are generated respectively. The time synchronization index and the sampling frequency difference index are predicted using a pre-trained machine learning model, and an association evaluation coefficient is generated based on the model. The association of data from different sources is intelligently predicted through the association evaluation coefficient.

4. The medical decision-making assistance method based on medical rules according to claim 2, characterized in that: The correlation evaluation coefficient generated by intelligently predicting the correlation of data from different sources through the machine learning model is compared and analyzed with the pre-set correlation evaluation coefficient reference threshold to further determine whether there is anomaly in data correlation. The specific steps are as follows: If the correlation evaluation coefficient is less than the correlation evaluation coefficient reference threshold, a correlation hidden danger signal is generated, indicating that data from different sources have the hidden danger of abnormal correlation; If the correlation evaluation coefficient is greater than or equal to the correlation evaluation coefficient reference threshold, an efficient correlation signal is generated, indicating that data from different sources can be efficiently correlated.

5. The medical decision-making assistance method based on medical rules according to claim 3 is characterized in that: After analyzing the time synchronization information of data from different devices, the specific steps for generating the time synchronization index are as follows: For each pair of associated data points from different devices, calculate the absolute difference in their timestamps. The timestamp of the data from device A is T A , the data timestamp of device B is T A , then the timestamp difference calculation formula is: D = |T A -T B |, where D represents the timestamp difference, reflecting the time interval between the two data points; The time difference weight factor is introduced to further quantify the impact of time difference. The time difference weight factor is inversely proportional to the timestamp difference. The calculation expression is: W D is the time difference weight factor, ranging from 0 to 0.5, which is used to measure the impact of timestamp differences. α is the first adjustment parameter, which controls the sensitivity of the weight factor to timestamp differences. Calculate the device weight factor to reflect the synchronization quality between devices. The device weight factor is calculated based on the known delay or transmission time of the device. The calculation expression is: W E is the device weight factor, ranging from 0 to 0.5, indicating the synchronization quality between devices, β is the second adjustment parameter, controlling the sensitivity of the device weight factor to latency, and L is the known latency or data transmission duration between devices; Comprehensive time difference weight factor W D and the device weight factor W E , calculate the time synchronization index, which reflects the overall synchronization degree of the data in time. The calculation expression is: TSI index =W D ·W E , TSI index It is the time synchronization index, which indicates the degree of time synchronization.

6. The medical decision-making assistance method based on medical rules according to claim 3, characterized in that: After analyzing the differences in data sampling frequencies of different devices, the specific steps for generating the sampling frequency difference index are as follows: Collect data sampling frequencies from different devices. Device A and device B are used to measure the patient’s physiological parameters, respectively. A and f B Respectively represent the sampling frequencies of device A and device B; Calculate the sampling frequency ratio between the two devices. The frequency ratio is calculated by the following formula: In the formula, R AB Indicates the sampling frequency ratio of device A and device B, max(f A , f B ) represents the highest sampling frequency of device A and device B, min(f A , f B ) represents the lowest sampling frequency of device A and device B; Calculate the absolute value of the sampling frequency difference between device A and device B to further understand the degree of frequency difference. The calculation expression for the absolute difference in sampling frequency is: AB =|f A -f B |, D AB Indicates the absolute difference in sampling frequency between device A and device B; The sampling frequency ratio R of device A and device B AB The absolute difference D between the sampling frequencies of device A and device B AB Calculate the sampling frequency difference index. The expression is: Where SFD index is the sampling frequency difference index, which indicates the degree of sampling frequency difference between devices, ln(1+R AB ) uses a logarithmic function to smooth the frequency ratio, The absolute value of the frequency difference is adjusted using a square root function to enhance the perception of the magnitude of the difference.

7. The medical decision-making assistance method based on medical rules according to claim 1, characterized in that: A clustering algorithm is used to conduct a comprehensive analysis of data from different sources. Based on the results of the clustering algorithm, data with abnormal correlations is identified. The specific steps are as follows: When anomalies in data association are detected, all relevant data points and their corresponding time synchronization index and sampling frequency difference index are collated; After the data is prepared, select a clustering algorithm to analyze the data; When performing cluster analysis, the data are grouped according to the input time synchronization index, sampling frequency difference index, and the standardized medical concept data embedding generated by mapping, thereby forming different cluster groups; Based on the results of the clustering algorithm, identify data points with abnormal associations; Data marked as abnormal were excluded from the final diagnosis and treatment decision.

8. The medical decision-making assistance method based on medical rules according to claim 7, characterized in that: Based on the results of the clustering algorithm, identify the data points with abnormal associations. The specific steps are as follows: Calculate the closeness within each group to assess the similarity between data points; The separation between groups was calculated to assess the differences between different groups; By comparing the internal compactness of the group and the separation between the groups, abnormal data points are identified. First, an upper threshold of the internal compactness and a lower threshold of the separation between the groups are set. For the internal compactness of the group, if the internal compactness of a data point or group exceeds the preset upper threshold, the data point is marked as an abnormal point; at the same time, if the separation between the data point or group and other groups is lower than the preset lower threshold, the data point is also marked as an abnormal point.

9. A medical decision support system based on medical rules, used to implement the medical decision support method based on medical rules as described in any one of claims 1 to 8, characterized in that: It includes data receiving module, data standardization and mapping module, related information acquisition module, related information analysis and screening module, intelligent prediction and anomaly detection module and anomaly analysis and processing module; A data receiving module receives the patient's medical data from various medical devices and information systems; The data standardization and mapping module standardizes the received data and maps the standardized data to the corresponding medical concepts after data standardization; The associated information acquisition module acquires and records the associated information of the data during the data mapping process; The correlation information analysis and screening module analyzes the obtained correlation information after obtaining it, preliminarily identifies potential mis-correlated data, and screens out the identified data to provide accurate input data for the prediction model; The intelligent prediction and anomaly detection module uses a pre-trained machine learning model to intelligently predict the association of data from different sources, and further determines whether there are anomalies in the data association based on the prediction results of the machine learning model; The abnormal analysis and processing module uses a clustering algorithm to conduct a comprehensive analysis of data from different sources when an abnormality is detected in data association. Based on the results of the clustering algorithm, the data with abnormal associations is identified, marked, and excluded from the final diagnosis and treatment decision.

Images