Intelligent prewarning system for perioperative period of high-risk surgical patient
By introducing a feedback mechanism from medical staff and dynamic priority correction rules into the traditional early warning system, the problem of rigid early warning priorities was solved, enabling adaptive adjustment of early warning priorities and system self-optimization, thereby improving the accuracy and timeliness of early warnings.
Patent Information
- Application Number
- CN202511686660.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-18
- Publication Date
- 2026-02-24
AI Technical Summary
Traditional early warning systems have rigid early warning priority settings that cannot adapt to the dynamic changes in patient risk during the perioperative period. This results in a serious disconnect between early warning priorities and actual clinical conditions, with important early warnings failing to be delivered in a timely manner, non-urgent early warnings interfering with normal medical work, and erroneous feedback data polluting system optimization.
By introducing a feedback mechanism from medical staff, dynamic priority correction rules based on confidence assessment are generated, and the warning priority is adjusted through an iterative optimization mechanism to ensure that the warning priority matches the actual clinical situation. This includes modules for information generation, push notification, feedback reception, data capture, score calculation, and priority adjustment.
It achieves adaptive adjustment of early warning priority, improves the accuracy of early warning, avoids system performance degradation, forms a virtuous cycle of self-evolution, and ensures timely delivery of early warning information and reasonable allocation of priorities.
Smart Images
Figure CN121565451A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of medical information technology, and in particular to an intelligent early warning system for high-risk surgical patients during the perioperative period. Background Technology
[0002] With the rapid development of medical information technology, intelligent early warning systems are playing an increasingly important role in the perioperative management of high-risk surgical patients. Traditional early warning systems typically assess and issue warnings about patient risks based on preset fixed rules. They monitor patients' vital signs and clinical indicators, triggering corresponding warning notifications when abnormal values are detected.
[0003] The above-disclosed technical solutions have at least the following technical problems: Traditional methods, due to their static rules for setting warning priorities, cannot adapt to the dynamic changes in perioperative patient risk, leading to a severe disconnect between warning priorities and actual clinical conditions. When surgical stages change or sudden complications occur, the system still classifies warning levels according to the original rules, resulting in high-risk warnings being classified as low to medium priority, while some routine risks are over-warned. This priority mismatch directly causes warning notification routing failures, preventing important warnings from reaching responsible medical staff in a timely manner, while a large number of non-urgent warnings interfere with normal medical work. More seriously, feedback data collected based on these erroneous warnings contaminates the system optimization process, causing the system to continuously reinforce incorrect directions, creating a vicious cycle. To address these problems, this invention proposes a solution. Summary of the Invention
[0004] This application provides an intelligent early warning system for high-risk surgical patients during the perioperative period, which solves the problem in the prior art where rigid early warning priority settings lead to a serious disconnect between system early warnings and actual clinical needs. It achieves the technical effects of dynamic adaptive adjustment of early warning priorities based on clinical feedback and continuous optimization of the system closed loop.
[0005] This application provides an intelligent early warning system for the perioperative period of high-risk surgical patients, including: an information generation module for generating initial early warning information based on the patient's perioperative data; Information push module: used to push the initial warning information to the medical staff's terminal, and provide priority adjustment feedback options when displayed on the terminal; Feedback receiving module: used to receive priority adjustment feedback submitted by medical staff through the terminal; Data capture module: used to capture data context related to the warning information in response to the priority adjustment feedback; Score calculation module: used to generate dynamic priority correction rules based on multiple priority adjustment feedbacks and their corresponding data contexts, and calculate a confidence score for each rule; Priority adjustment module: used to apply the dynamic priority correction rule to match the data context and adjust the priority of the new warning information when generating new warning information; Iterative optimization module: used to iteratively optimize the dynamic priority correction rule based on priority adjustment feedback.
[0006] Furthermore, the steps for generating initial warning information based on patient perioperative data include: The risk assessment model is used to analyze patients' preoperative, intraoperative and postoperative data and output a risk score. The risk score is mapped to an initial priority based on a preset threshold, and the initial priority includes three levels: high, medium, and low.
[0007] Furthermore, the steps of generating dynamic priority correction rules and calculating confidence scores include: Frequent itemsets are extracted from historical feedback data using association rule mining algorithms, and candidate rules are formed by associating frequent itemsets with feedback directions. For each candidate rule R: if the adjustment priority of condition C changes from X to Y, its confidence score is... The calculation formula is: ; in, This represents the number of feedback samples in historical feedback data that match condition C and whose feedback direction is consistent with the adjustment direction. This represents the total number of responses to matching condition C. This represents the variance of the feedback direction value of the matching condition C. To adjust the parameters; The condition C is defined by a combination of features in the data context, including vital sign abnormality indicators, surgical stage codes, and medical operation types. The confidence score Used to assess the reliability of the rules. The higher the value, the more reliable the rule.
[0008] Furthermore, the steps for adjusting the priority of new early warning information using dynamic priority correction rules include: For a new early warning message, with data context X and initial priority P, calculate the priority adjustment value. : ; in, To match the dynamic priority correction rules of data context X, For rules The adjustment value is +1 for increasing and -1 for decreasing. For rules The weight is calculated using the following formula: ; in, For rules The confidence score, For data context X and rule conditions The matching distance; Final adjusted priority However, this is limited to a specific priority level.
[0009] Furthermore, the steps for iteratively optimizing the dynamic priority correction rule include: When applying rules When the adjusted early warning receives subsequent priority adjustment feedback, it will be adjusted according to the direction of the feedback. With rules and actions Consistency, update rules confidence score : ; in, For learning rate, This is an indicator function; it is 1 when the new feedback direction matches the rule action, and 0 otherwise. if If the rule falls below the elimination threshold, it is removed from the rule base. .
[0010] Furthermore, the steps for generating dynamic priority correction rules include: Frequent itemsets are extracted from historical feedback data using an association rule mining algorithm. These frequent itemsets are composed of feature conditions in the data context. Frequent itemsets are associated with feedback directions to form candidate rules. Then, a confidence score for each candidate rule is calculated using a confidence formula. Only rules with confidence scores higher than the minimum threshold are retained.
[0011] Furthermore, the steps for the risk assessment model to analyze the patient's preoperative, intraoperative, and postoperative data are as follows: Constructing a multi-dimensional risk assessment matrix Its elements are determined by the coefficient of variation of physiological parameters within different time windows, the risk coefficient of the surgical stage, and the correlation with complications; The formula for calculating the risk assessment matrix is as follows: ; in, Indicates the first The physiological parameter at the first coefficient of variation within a time window Indicates the first Each time window corresponds to a risk factor in the surgical stage. Indicates the first The correlation between physiological parameters and complications , , These are the weighting coefficients, and ; Risk coefficient through matrix Frobenius norm calculation: .
[0012] Furthermore, in response to the priority adjustment feedback, the step of capturing the data context related to the warning information includes: Establish multidimensional data context vectors Its dimensions include vital sign trend components, surgical stage coding components, concurrency warning status components, and medical operation record components; The vital signs trend component The calculation formula is: ; in, Indicates the first A function of how vital signs change over time. Its derivative, To monitor the time window, For the first Weighting coefficients for each vital sign parameter; This represents the total number of vital sign parameters. The surgical phase coding components One-hot encoding is used to represent the current surgical stage and the concurrent warning status component. Other simultaneous warning states in the system are represented by binary vectors, and the medical operation record component is also included. The importance of recent medical procedures is represented by a weighted frequency vector; Final Data Context Vector This constitutes a complete contextual representation.
[0013] One or more technical solutions provided in the embodiments of this application have at least the following technical effects or advantages: By introducing a dynamic feedback mechanism for early warning priorities from medical staff, the system can automatically adjust early warning priorities based on actual clinical conditions, solving the problem of rigid priority settings in the original system. Furthermore, during the priority adjustment process, by capturing complete clinical data context, including vital sign trends and surgical stages, the system ensures that priority adjustments are based on sufficient clinical evidence, improving the accuracy of early warnings. Even further, by establishing a rule-based evaluation system and iterative optimization mechanism based on confidence scores, the system can continuously learn and improve, avoiding system performance degradation caused by erroneous early warnings and forming a virtuous cycle of self-evolution. Attached Figure Description
[0014] Figure 1 This is a schematic diagram of the structure of an intelligent early warning system for high-risk surgical patients during the perioperative period, provided in an embodiment of this application. Detailed Implementation
[0015] This application provides an intelligent early warning system for high-risk surgical patients during the perioperative period, which solves the problem of rigid early warning priority settings in the prior art leading to a disconnect between early warnings and clinical reality. By establishing a feedback mechanism for medical staff, generating dynamic rules based on confidence assessment, and continuously iteratively optimizing, the system achieves the technical effects of adaptive adjustment of early warning priorities and self-evolution of the system.
[0016] To better understand the above technical solutions, the following will provide a detailed explanation of the technical solutions in conjunction with the accompanying drawings and specific implementation methods.
[0017] like Figure 1 As shown in the figure, this application provides a perioperative intelligent early warning system for high-risk surgical patients, including: an information generation module: used to generate initial early warning information based on the patient's perioperative data, wherein the initial early warning information includes early warning content and initial priority; Information push module: used to push the initial warning information to the medical staff's terminal, and provide priority adjustment feedback options when displayed on the terminal; Feedback receiving module: used to receive priority adjustment feedback submitted by medical staff through the terminal; Data capture module: In response to the priority adjustment feedback, it captures the data context related to the warning information. The data context includes the patient's real-time vital sign data trend, the precise stage of the surgery, other warning information existing in the system at the same time, and recent important medical operations and medication records. The score calculation module is used to generate dynamic priority correction rules based on multiple priority adjustment feedbacks and their corresponding data contexts by analyzing common data features, and to calculate a confidence score for each rule. Priority adjustment module: used to apply the dynamic priority correction rule to match the data context and adjust the priority of the new warning information when generating new warning information; Iterative optimization module: used to iteratively optimize the dynamic priority correction rule based on priority adjustment feedback.
[0018] Furthermore, the steps for generating initial warning information based on patient perioperative data include: The risk assessment model is used to analyze patients' preoperative, intraoperative and postoperative data and output a risk score. The risk score is mapped to an initial priority based on a preset threshold, and the initial priority includes three levels: high, medium, and low.
[0019] Furthermore, the steps of generating dynamic priority correction rules and calculating confidence scores include: Frequent itemsets are extracted from historical feedback data using an association rule mining algorithm. These frequent itemsets are composed of feature conditions in the data context, and the frequent itemsets are associated with the feedback direction to form candidate rules. For each candidate rule R: if the adjustment priority of condition C changes from X to Y, its confidence score is... The calculation formula is: ; in, This represents the number of feedback samples in historical feedback data that match condition C and whose feedback direction is consistent with the adjustment direction. This represents the total number of responses to matching condition C. This represents the variance of the feedback direction value of the matching condition C. The parameters are adjusted to penalize inconsistencies in direction; The condition C is defined by a combination of features in the data context, including vital sign abnormality indicators, surgical stage codes, and medical operation types. The confidence score Used to assess the reliability of the rules. The higher the value, the more reliable the rule.
[0020] The steps for using the formula are as follows: First, filter out all feedback records that match condition C from the historical feedback data, and then calculate... and Then calculate the variance of the feedback direction value. Finally, substitute into the formula to calculate. Among them, the feedback direction value The variance is +1 (indicating "higher priority") or -1 (indicating "lower priority"), , This is the mean value of the feedback direction for matching condition C.
[0021] Furthermore, the steps for adjusting the priority of new early warning information using dynamic priority correction rules include: For a new early warning message, with data context X and initial priority P, calculate the priority adjustment value. : ; in, To match the dynamic priority correction rules of data context X, For rules The adjustment value is +1 for increasing and -1 for decreasing. For rules The weight is calculated using the following formula: ; in, For rules The confidence score, For data context X and rule conditions The matching distance indicates the degree of mismatch; Final adjusted priority However, this is limited to a specific priority range; The matching distance By calculating X and The result is obtained by normalizing the Euclidean distance in the feature space.
[0022] The steps for using the formula are as follows: First, combine the data context X and the rule conditions. Convert the numerical values into feature vectors and calculate the Euclidean distance. ,in and X and The j-th eigenvalue is then normalized. , The preset maximum distance value is used; the final calculation is then performed. and And adjust the priority.
[0023] Furthermore, the steps for iteratively optimizing the dynamic priority correction rule include: When applying rules When the adjusted early warning receives subsequent priority adjustment feedback, it will be adjusted according to the direction of the feedback. With rules and actions Consistency, update rules confidence score : ; in, For learning rate, This is an indicator function; it is 1 when the new feedback direction matches the rule action, and 0 otherwise. if If the rule falls below the elimination threshold, it is removed from the rule base. ; The steps for using the formula are as follows: Whenever there is new feedback Upon arrival, check its compliance with the rules of action. To ensure consistency, calculate the indicator function value and then substitute it into the update formula for adjustment. Among them, the learning rate The default value is 0.1, and the elimination threshold is set to 0.5.
[0024] Furthermore, the steps for generating dynamic priority correction rules include: Frequent itemsets are extracted from historical feedback data using an association rule mining algorithm. These frequent itemsets are composed of feature conditions in the data context. Frequent itemsets are associated with feedback directions to form candidate rules. Then, a confidence score for each candidate rule is calculated using a confidence formula. Only rules with confidence scores higher than the minimum threshold are retained.
[0025] Furthermore, the steps for the risk assessment model to analyze the patient's preoperative, intraoperative, and postoperative data are as follows: Constructing a multi-dimensional risk assessment matrix Its elements are determined by the coefficient of variation of physiological parameters within different time windows, the risk coefficient of the surgical stage, and the correlation with complications; The formula for calculating the risk assessment matrix is as follows: ; in, Indicates the first The physiological parameter at the first coefficient of variation within a time window Indicates the first Each time window corresponds to a risk factor in the surgical stage. Indicates the first The correlation between physiological parameters and complications , , These are the weighting coefficients, and ; Risk coefficient through matrix Frobenius norm calculation: ; The steps for using the formula are as follows: First, collect physiological parameter data according to the time window sequence, and calculate the coefficient of variation of each parameter in different time windows. Then obtain the surgical stage checklist. Search from the complication knowledge base Finally, substitute the values into the formula to calculate the risk score. ; where the coefficient of variation , and They represent the first The physiological parameter at the first The standard deviation and mean within each time window.
[0026] Furthermore, in response to the priority adjustment feedback, the step of capturing the data context related to the warning information includes: Establish multidimensional data context vectors Its dimensions include vital sign trend components, surgical stage coding components, concurrency warning status components, and medical operation record components; The vital signs trend component The calculation formula is: ; in, Indicates the first A function of how vital signs change over time. Its derivative, To monitor the time window, For the first Weighting coefficients for each vital sign parameter; This represents the total number of vital sign parameters. The surgical phase coding components One-hot encoding is used to represent the current surgical stage and the concurrent warning status component. Other simultaneous warning states in the system are represented by binary vectors, and the medical operation record component is also included. The importance of recent medical procedures is represented by a weighted frequency vector; Final Data Context Vector This constitutes a complete contextual representation.
[0027] The steps for using the formula are as follows: First, acquire continuous time series data from the vital signs monitoring device, calculate the integral of the rate of change of each parameter through numerical differentiation, and then combine the surgical progress, concurrency warning status and medical operation records to synthesize a multidimensional data context vector according to preset weights; the integral of the rate of change is calculated using the trapezoidal numerical integration method, and the weight of medical operation is preset according to the criticality of the operation type.
[0028] The above formulas are all dimensionless calculations. The formulas are derived from software simulations based on a large amount of collected data to obtain the most recent real-world results. The preset parameters in the formulas are set by those skilled in the art according to the actual situation.
[0029] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, in the form of a computer program product.
[0030] Those skilled in the art will recognize that the modules and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.
[0031] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.
[0032] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.
[0033] In conclusion, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.
Claims
1. A perioperative intelligent early warning system for high-risk surgical patients, characterized in that, include: Information generation module: used to generate initial warning information based on the patient's perioperative data; Information push module: used to push the initial warning information to the medical staff's terminal, and provide priority adjustment feedback options when displayed on the terminal; Feedback receiving module: used to receive priority adjustment feedback submitted by medical staff through the terminal; Data capture module: used to capture data context related to the warning information in response to the priority adjustment feedback; Score calculation module: used to generate dynamic priority correction rules based on multiple priority adjustment feedbacks and their corresponding data contexts, and calculate a confidence score for each rule; Priority adjustment module: used to apply the dynamic priority correction rule to match the data context and adjust the priority of the new warning information when generating new warning information; Iterative optimization module: used to iteratively optimize the dynamic priority correction rule based on priority adjustment feedback.
2. The perioperative intelligent early warning system for high-risk surgical patients as described in claim 1, characterized in that, The steps for generating initial warning information based on patient perioperative data include: The risk assessment model is used to analyze patients' preoperative, intraoperative and postoperative data and output a risk score. The risk score is mapped to an initial priority based on a preset threshold, and the initial priority includes three levels: high, medium, and low.
3. The perioperative intelligent early warning system for high-risk surgical patients as described in claim 1, characterized in that, The steps for generating dynamic priority correction rules and calculating confidence scores include: Frequent itemsets are extracted from historical feedback data using association rule mining algorithms, and candidate rules are formed by associating frequent itemsets with feedback directions. For each candidate rule R: if the adjustment priority of condition C changes from X to Y, its confidence score is... The calculation formula is: ； in, This represents the number of feedback samples in historical feedback data that match condition C and whose feedback direction is consistent with the adjustment direction. This represents the total number of responses to matching condition C. This represents the variance of the feedback direction value of the matching condition C. To adjust the parameters; The condition C is defined by a combination of features in the data context, including vital sign abnormality indicators, surgical stage codes, and medical operation types. The confidence score Used to assess the reliability of the rules. The higher the value, the more reliable the rule.
4. The perioperative intelligent early warning system for high-risk surgical patients as described in claim 3, characterized in that, The steps for adjusting the priority of new warning information using dynamic priority correction rules include: For a new early warning message, with data context X and initial priority P, calculate the priority adjustment value. : ； in, To match the dynamic priority correction rules of data context X, For rules The adjustment value is +1 for increasing and -1 for decreasing. For rules The weight is calculated using the following formula: ； in, For rules The confidence score, For data context X and rule conditions The matching distance; Final adjusted priority However, this is limited to a specific priority level.
5. The perioperative intelligent early warning system for high-risk surgical patients as described in claim 4, characterized in that, The steps for iteratively optimizing the dynamic priority correction rule include: When applying rules When the adjusted early warning receives subsequent priority adjustment feedback, it will be adjusted according to the direction of the feedback. With rules and actions Consistency, update rules confidence score : ； in, For learning rate, This is an indicator function; it is 1 when the new feedback direction matches the rule action, and 0 otherwise. if If the rule falls below the elimination threshold, it is removed from the rule base. .
6. The perioperative intelligent early warning system for high-risk surgical patients as described in claim 1, characterized in that, The steps for generating dynamic priority correction rules include: Frequent itemsets are extracted from historical feedback data using an association rule mining algorithm. These frequent itemsets are composed of feature conditions in the data context. Frequent itemsets are associated with feedback directions to form candidate rules. Then, a confidence score for each candidate rule is calculated using a confidence formula. Only rules with confidence scores higher than the minimum threshold are retained.
7. The perioperative intelligent early warning system for high-risk surgical patients as described in claim 6, characterized in that, The steps for the risk assessment model to analyze preoperative, intraoperative, and postoperative patient data are as follows: Constructing a multi-dimensional risk assessment matrix Its elements are determined by the coefficient of variation of physiological parameters within different time windows, the risk coefficient of the surgical stage, and the correlation with complications; The formula for calculating the risk assessment matrix is as follows: ； in, Indicates the first The physiological parameter at the first coefficient of variation within a time window Indicates the first Each time window corresponds to a risk factor in the surgical stage. Indicates the first The correlation between physiological parameters and complications , , These are the weighting coefficients, and ; Risk coefficient through matrix Frobenius norm calculation: 。 8. The perioperative intelligent early warning system for high-risk surgical patients as described in claim 1, characterized in that, In response to the priority adjustment feedback, the step of capturing the data context related to the warning information includes: Establish multidimensional data context vectors Its dimensions include vital sign trend components, surgical stage coding components, concurrency warning status components, and medical operation record components; The vital signs trend component The calculation formula is: ； in, Indicates the first A function of how vital signs change over time. Its derivative, To monitor the time window, For the first Weighting coefficients for each vital sign parameter; This represents the total number of vital sign parameters. The surgical phase coding components One-hot encoding is used to represent the current surgical stage and the concurrent warning status component. Other simultaneous warning states in the system are represented by binary vectors, and the medical operation record component is also included. The importance of recent medical procedures is represented by a weighted frequency vector; Final Data Context Vector This constitutes a complete contextual representation.