Monitoring information analysis system for surgical procedures

Abstract

This invention relates to the field of hypotension risk pattern prediction technology, specifically to a monitoring information analysis system for surgical procedures. The invention first acquires blood pressure monitoring data and obtains the analysis interval for each intraoperative event; further, it analyzes the hypotension manifestation of blood pressure in each event to obtain the blood pressure abnormality degree; further, based on the distribution of blood pressure abnormality degree in historical patient data for each hypotension impact stage, it obtains the stage-specific abnormality manifestation degree and integrates it to obtain event attention; further, it acquires the latest real-time operational event set of the current patient, and in the most matching hypotension impact stage, based on the similarity characteristics between the real-time operational event set and the surgical operation in each hypotension impact stage, combined with the stage-specific abnormality manifestation degree and event attention, it obtains the operation-blood pressure linkage impact degree, and uses a neural network model for real-time early warning, solving the problem of early warning lag caused by ignoring the impact of operations and individual differences, thus ensuring patient safety.

Description

Technical Field

[0001] This invention relates to the field of low blood pressure risk pattern prediction technology, and more specifically to a monitoring information analysis system for surgical procedures. Background Technology

[0002] Low blood pressure can lead to insufficient blood supply to vital organs such as the heart, brain, and kidneys, resulting in organ hypoxia, functional failure, and in severe cases, even death. Therefore, real-time monitoring and early warning of low blood pressure are particularly important in surgical procedures.

[0003] Currently, during cardiac surgery, neural networks are generally used to monitor and analyze the patient's physiological parameters (such as blood pressure, heart rate, and temperature) to provide early warnings of intraoperative hypotension. However, due to the lack of correlation analysis between the surgeon's actions during surgery and intraoperative hypotension, traditional prediction methods have a certain lag and cannot provide timely warnings immediately after the surgery, making patients susceptible to the risk of intermittent hypotension during surgery. Summary of the Invention

[0004] To address the technical problem that existing intraoperative hypotension early warning methods cannot predict the impact of intraoperative procedures in a timely manner, resulting in a certain lag in early warning, the present invention aims to provide a monitoring information analysis system for surgical procedures. The specific technical solution adopted is as follows:

[0005] A monitoring information analysis system for surgical procedures, the system comprising:

[0006] Monitoring module: acquires monitoring data for three preset blood pressure values; annotates surgical procedures at various stages during the operation and divides the analysis interval for each event;

[0007] Analysis module: Based on the low-pressure manifestation of the three blood pressure types within the analysis interval of each event, obtain the blood pressure abnormality degree; in the historical patient data, divide the low blood pressure impact stage according to the fluctuation of the blood pressure abnormality degree; classify the low blood pressure impact stage according to the similarity characteristics of surgical operations between different low blood pressure impact stages; obtain the stage-specific abnormality performance degree of each low blood pressure impact stage according to the distribution of the blood pressure abnormality degree of each category of low blood pressure impact stage, and merge them to obtain the event attention degree;

[0008] Real-time early warning module: Constructs a real-time operation event set by including all surgical operations within a preset historical neighborhood of the patient's latest surgical operation; In the hypotension impact stage that best matches the real-time operation event set, Based on the similarity characteristics between the real-time operation event set and the surgical operations of each hypotension impact stage, combined with the stage-specific abnormality performance and the event attention, the operation-blood pressure linkage impact degree is obtained, and real-time early warning is provided with the help of a neural network model.

[0009] Furthermore, the method for obtaining the degree of blood pressure abnormality includes:

[0010] Within the analysis interval of each event, each type of blood pressure is taken as the target blood pressure, and data where the target blood pressure is lower than the preset diastolic reference threshold are marked as abnormal data. Based on the time proportion of the abnormal data and the area between the data curve of the abnormal data and the reference line of the preset diastolic reference threshold, the diastolic abnormality of the target blood pressure is obtained.

[0011] By integrating the low-pressure abnormality manifestations of the three blood pressure data, the blood pressure abnormality of the corresponding event is obtained.

[0012] Furthermore, the method for fusing the low-pressure abnormality manifestation of the three blood pressure values ​​to obtain the blood pressure abnormality of the corresponding event includes:

[0013] For each event, the normalized result of the low-pressure abnormality of the mean arterial pressure is used as the first weight, the difference between constant 1 and the first weight is used as the second weight, the low-pressure abnormality of the systolic pressure is weighted based on the first weight, the low-pressure abnormality of the diastolic pressure is weighted based on the second weight, and the weighted sum is used as the blood pressure abnormality of the corresponding event.

[0014] Furthermore, the method for classifying the stages of hypotension includes:

[0015] Based on the intersection-union ratio of surgical procedures between any two hypotension-affected stages, the clustering distance between the corresponding two hypotension-affected stages is obtained. Clustering is performed based on all the clustering distances, and each cluster corresponds to a classification of a hypotension-affected stage.

[0016] Furthermore, the method for obtaining event attention includes:

[0017] In each of the hypotension impact stages, the abnormal impact span is obtained based on the fluctuation range of the blood pressure abnormality in each hypotension impact stage; the abnormal impact span of each hypotension impact stage and the maximum blood pressure abnormality are combined to obtain the stage abnormality performance.

[0018] Event attention is obtained based on the overall characteristics of all the stage-specific abnormal manifestations in each of the hypotension impact stages.

[0019] Furthermore, the method for obtaining the operation-blood pressure linkage effect includes:

[0020] In the hypotension-affected stage that best matches the real-time operation event set, a comparison reference weight is obtained based on the similarity characteristics between the real-time operation event set and the surgical operation of each hypotension-affected stage.

[0021] By integrating the comparative reference weights and the corresponding stage-specific abnormality of the hypotension impact stage, the local linkage impact degree of the real-time operation event set and each hypotension impact stage is obtained.

[0022] Based on the overall characteristics of the local linkage influence, and combined with the event attention, the operation-blood pressure linkage influence is obtained.

[0023] Furthermore, the method for obtaining the comparison reference weights includes:

[0024] The Jaccard correlation coefficient between the surgical operations of the real-time operation event set and the surgical operations of each of the hypotension-affected stages is used as a comparative reference weight.

[0025] Furthermore, the most matching class of hypotension impact stages in the real-time operation event set is obtained according to the nearest neighbor algorithm.

[0026] Furthermore, the method for obtaining the hypotension effect phase includes:

[0027] In the data of each historical patient, the time domain is divided into different stages of hypotension influence by using the maximum value of the blood pressure abnormality in the time domain as the dividing point.

[0028] Furthermore, the method for obtaining the analysis interval includes:

[0029] The time interval between the start time of each surgical procedure and the start time of the next surgical procedure is used as the analysis interval for each event.

[0030] The present invention has the following beneficial effects:

[0031] This invention first acquires blood pressure monitoring data and obtains the analysis interval for each intraoperative event, providing a basis for subsequent analysis. It further analyzes the diastolic blood pressure manifestation of each event, obtaining the blood pressure abnormality degree to characterize the degree of diastolic blood pressure abnormality after a single event, providing a basis for subsequent segmentation into different stages. Further, it segments and classifies the hypotension impact stages from historical patient data, extracting different typical, repeatable known dangerous operation patterns from historical data. Based on the distribution of blood pressure abnormality degree in each hypotension impact stage, it obtains the stage-specific abnormality degree for each hypotension impact stage, quantifies the risk level of each hypotension impact stage, and smooths individual differences by fusing the stage-specific abnormality degree to obtain event attention, making the prediction more universal and robust for new patients. Finally, it acquires the latest real-time operation event set for the current patient. In the most matching hypotension impact stage, based on the similarity characteristics between the real-time operation event set and the surgical operation in each hypotension impact stage, combined with the stage-specific abnormality degree and event attention, it reflects the impact of the surgical operation on blood pressure from both local and global perspectives, obtaining the operation-blood pressure linkage impact degree, providing more reference for the neural network model, and finally using the neural network model for real-time early warning. The solution dynamically quantifies the potential impact of surgical procedures by linking real-time operational events with the most matching hypotension impact stage, and inputs the data into a neural network to achieve early warning. This solves the problem of delayed warning caused by ignoring the impact of procedures and individual differences, thus ensuring patient safety. Attached Figure Description

[0032] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0033] Figure 1 A system block diagram of a monitoring information analysis system for surgical procedures provided in one embodiment of the present invention;

[0034] Figure 2 A data curve of systolic blood pressure is provided as an embodiment of the present invention;

[0035] Figure 3 This is a flowchart of a method for obtaining the influence of the operation-blood pressure linkage, provided as an embodiment of the present invention. Detailed Implementation

[0036] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a monitoring information analysis system for surgical procedures proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0037] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0038] The following description, in conjunction with the accompanying drawings, details a specific solution for a monitoring information analysis system for surgical procedures provided by the present invention.

[0039] Please see Figure 1 The diagram illustrates a system block diagram of a monitoring information analysis system for surgical procedures according to an embodiment of the present invention. The system includes a monitoring module 101, an analysis module 102, and a real-time early warning module 103.

[0040] The monitoring module 101 is used to acquire monitoring data of three preset blood pressure values; to annotate the surgical operations of each stage during the operation, and to divide the analysis interval of each event.

[0041] Different types of surgeries involve different physiological systems, and the intervention methods during surgery also vary. These factors collectively determine the risk of intraoperative hypotension. Especially in cardiac surgery, due to the direct manipulation of the heart and vascular system, the frequency and severity of intraoperative hypotension are relatively high. Therefore, this invention provides a precise intraoperative hypotension early warning analysis for cardiac surgery, thereby acquiring monitoring data for three preset blood pressure levels.

[0042] In cardiac surgery, hypotension is often the result of a combination of factors, and the surgeon's actions at each stage of the procedure can have an impact. First, the choice and dosage of anesthetic drugs directly affect vasoconstriction and vasodilation; excessive anesthetic drugs can inhibit the sympathetic nervous system, leading to a drop in blood pressure. Changes in patient position during surgery, especially changes in blood return after the chest cavity is opened, can also trigger hypotension. In addition, blood volume loss caused by bleeding, cardiac manipulation, or improper medication management during surgery can further exacerbate hypotension.

[0043] Therefore, by annotating the surgical procedures at each stage of the operation and dividing each event into analysis intervals, detailed analysis and early warning can be facilitated. For a complete cardiac surgery process, it can be roughly divided into several stages, and each stage corresponds to relevant operational events:

[0044] (1) Preoperative procedures: anesthesia induction, endotracheal intubation, central venous catheterization, etc.;

[0045] (2) Extracorporeal circulation related procedures: aortic cannulation, vena cava cannulation, CPB initiation, aortic clamping, etc.;

[0046] (3) Key events in cardiac procedures: cardiac tumbling / lifting, coronary anastomosis (OPCAB), mitral valve retractor traction, etc.;

[0047] (4) Major vascular procedures: aortic sidewall clamping, proximal aortic anastomosis (CABG), PDA ligation, etc.;

[0048] (5) Medications and special interventions: protamine injection, cardiac defibrillation, etc.;

[0049] (6) Postoperative procedures: chest closure / sternal wire fixation, body position changes (head elevated), endotracheal tube suction, etc.

[0050] Each step in each stage is considered a surgical procedure and is recorded as an operational event, or simply an event.

[0051] In one embodiment of the present invention, the time domain interval between the start time of each surgical operation and the start time of the next surgical operation is used as the analysis interval for each event.

[0052] It acquires continuous blood pressure monitoring data from doctors during procedures, with three preset blood pressure values: systolic pressure, diastolic pressure, and mean arterial pressure.

[0053] Among them, diastolic blood pressure directly affects coronary artery perfusion during diastole, because myocardial blood supply mainly occurs during diastole. Low diastolic blood pressure can lead to an increased risk of myocardial ischemia, which is crucial for maintaining the balance between myocardial oxygen supply and demand during cardiac surgery. Systolic blood pressure reflects cardiac output and myocardial contractility. A decrease in systolic blood pressure may indicate heart failure or insufficient blood volume, affecting the blood flow impact and perfusion efficiency of organs throughout the body (such as the brain and kidneys). Mean arterial pressure, on the other hand, represents the average perfusion pressure throughout the cardiac cycle. It integrates the pressure changes during systole and diastole and is the core determinant of organ perfusion pressure. Mean arterial pressure below the threshold is usually significantly associated with postoperative myocardial injury, kidney injury, and mortality risk.

[0054] It should be noted that the three methods of blood pressure collection are well-known technologies, and the collection frequency can be set by the implementer according to the implementation scenario, so they will not be elaborated further.

[0055] Analysis module 102 is used to obtain the blood pressure abnormality degree based on the low blood pressure manifestation of the three blood pressure types within the analysis interval of each event; to divide the low blood pressure impact stage in historical patient data according to the fluctuation of blood pressure abnormality degree; to classify the low blood pressure impact stage according to the similarity characteristics of surgical operations between different low blood pressure impact stages; and to obtain the stage-specific abnormality manifestation degree of each low blood pressure impact stage and merge them to obtain the event attention degree based on the distribution of blood pressure abnormality degree of each type of low blood pressure impact stage.

[0056] After marking the analysis intervals, we can analyze the abnormal blood pressure performance after each surgical event. This embodiment of the invention mainly analyzes diastolic abnormalities. Therefore, based on the diastolic performance of the three types of blood pressure within the analysis interval of each event, we obtain the degree of blood pressure abnormality, which characterizes the degree of diastolic abnormality of blood pressure after a single event, and provides a basis for subsequent division of different stages.

[0057] Preferably, in one embodiment of the present invention, considering that each blood pressure has its own medical data reference standard, the preset low pressure reference thresholds for systolic blood pressure, diastolic blood pressure and mean arterial pressure are set to 90, 60 and 65 mmHg, respectively; within the analysis interval of each event, each blood pressure is taken as the target blood pressure, and data with target blood pressure lower than the preset low pressure reference threshold are marked as abnormal data.

[0058] Considering that the larger the proportion of abnormal data in time, the higher the degree of abnormality from a time perspective; when the area between the abnormal data curve and the reference line of the preset blood pressure reference threshold is larger, it indicates that the severity of hypotension and the cumulative physiological damage load are higher. Therefore, based on the proportion of abnormal data in time, combined with the area between the abnormal data curve and the reference line of the preset blood pressure reference threshold, the degree of hypotension abnormality of the target blood pressure is obtained.

[0059] As an example: Please refer to Figure 2 It shows a data curve of systolic blood pressure provided by an embodiment of the present invention. Figure 2 The horizontal axis represents the time axis, with units in seconds, and the vertical axis represents the data axis, with units in mmHg. The solid curve represents the systolic blood pressure data curve, and the horizontal dashed line represents the reference line for the preset blood pressure reference threshold. A and B are the starting and ending points of abnormal data, respectively. The area between the AB curve segment and the AB dashed line segment is obtained by integration, and the area is linearly normalized in the corresponding data dimension.

[0060] The proportion of the duration of abnormal data to the total duration of the analysis interval is used as the time percentage. The product of the time percentage and the corresponding area is used as the degree of diastolic abnormality of the target blood pressure in the corresponding analysis interval. The diastolic performance of the target blood pressure is represented by the time percentage and the area corresponding to the abnormal data.

[0061] In other embodiments of the present invention, the implementer may also use a weighted summation method to combine the time proportion and the corresponding area, such as weights of 0.6 and 0.4 respectively.

[0062] Furthermore, by integrating the diastolic abnormality of the three blood pressure levels, the blood pressure abnormality of the corresponding event can be obtained.

[0063] As an example: Considering the early warning process of hypotension, when the mean arterial pressure is more abnormal, the diastolic pressure should be given more attention, because a severe decrease in mean arterial pressure directly threatens coronary artery perfusion (while myocardial blood supply mainly depends on the heart's diastolic phase), thus significantly increasing the risk of insufficient organ perfusion (such as postoperative myocardial injury, kidney injury); while when the mean arterial pressure is slightly abnormal or normal, the systolic pressure should be given priority, because a normal mean arterial pressure may mask an isolated abnormality in systolic pressure, and systolic pressure reflects cardiac output and myocardial contractility. A decrease in systolic pressure suggests insufficient cardiac pump function or hypovolemia;

[0064] Based on this, for each event, the normalized result of the low-pressure abnormality of mean arterial pressure is used as the first weight, and the difference between constant 1 and the first weight is used as the second weight. The low-pressure abnormality of systolic blood pressure is weighted based on the first weight, and the low-pressure abnormality of diastolic blood pressure is weighted based on the second weight. The weighted sum is used as the blood pressure abnormality of the corresponding event, reflecting the low-pressure manifestation of the three blood pressures.

[0065] Specifically, the formula for calculating the degree of blood pressure abnormality includes: Q represents the degree of blood pressure abnormality of the event being analyzed; W represents the first weight of the event being analyzed. The second weight of the event being analyzed is indicated by E, which represents the degree of abnormality in systolic blood pressure of the event being analyzed, and R, which represents the degree of abnormality in diastolic blood pressure of the event being analyzed.

[0066] It should be noted that the analysis process for the degree of diastolic abnormality for each blood pressure level is consistent for each event, and the analysis process for the degree of blood pressure abnormality is consistent for each event; only one example is described here.

[0067] Considering that the effects of hypotension caused by certain procedures may not be immediately apparent, but rather accumulate gradually through multiple stages of the procedure and only manifest when a specific event occurs, assessing the impact of hypotension requires considering the synergistic effects of each stage and comprehensively analyzing the interrelationships and potential cumulative effects between events to more accurately predict and warn of hypotension.

[0068] Therefore, based on the fluctuations in blood pressure abnormalities in historical patient data, the hypotension impact stages are divided. Then, based on the similar characteristics of surgical procedures between different hypotension impact stages, the hypotension impact stages are classified. Each category represents a typical, repeatable dangerous operation pattern. Valuable clinical knowledge is extracted from historical data to provide a reliable predictive basis for subsequent real-time prediction.

[0069] Preferably, in one embodiment of the present invention, considering that after a set of operations begins, blood pressure begins to be cumulatively affected, and the blood pressure abnormality gradually reaches its peak, and then the influence may begin to change due to medical intervention, the end of the operation or the entry into a new stage, the time domain is divided into different hypotension influence stages in the data of each historical patient, with the maximum value of blood pressure abnormality in the time domain as the dividing point.

[0070] Preferably, in one embodiment of the present invention, considering that the greater the proportion of identical surgical procedures between different hypotension-affected stages, the more similar the two stages are and the smaller the clustering distance is, the similarity characteristics of surgical procedures between different hypotension-affected stages are shown based on the crossover and union ratio of surgical procedures between any two hypotension-affected stages, thereby obtaining the clustering distance between the corresponding two hypotension-affected stages.

[0071] As an example, the crossover ratio of surgical procedures between any two hypotension-affected stages is increased by a positive parameter of 0.01 (divided by zero), and then the reciprocal is taken as the cluster distance between the corresponding two hypotension-affected stages.

[0072] Further clustering was performed based on the distance between all clusters, with each cluster corresponding to a classification of the hypotension-affected stage.

[0073] As an example, clustering classification is performed using DBSCAN clustering.

[0074] It should be noted that in other embodiments of the present invention, the crossover ratio can also be negatively correlated using a negative correlation function, such as the negative exponential function exp(-x) with the natural constant e as the base, to obtain the clustering distance, where x is the independent variable; other clustering algorithms such as K-means clustering can also be used for classification, all of which are well-known techniques and will not be described in detail here.

[0075] For each stage of hypotension, although there are some similar operational events, the physiological response to the same surgical procedure may vary due to factors such as the patient's physiological characteristics, drug response, and preoperative risk assessment results. At the same time, the skill level, operating habits, judgment of the intraoperative situation, and choice of treatment interventions of different doctors will also cause the same event to show different results in different situations, resulting in significant differences in the timing, severity, and duration of hypotension.

[0076] Therefore, it is necessary to analyze the degree of impact on hypotension presented by each type of similar operational event. Considering that the distribution of blood pressure abnormality reflects the severity of blood pressure instability and the concentration trend of blood pressure risk within the stage, the stage-specific abnormality performance of each hypotension impact stage is obtained based on the distribution of blood pressure abnormality in each hypotension impact stage. The risk level of each hypotension impact stage is quantified, and individual differences are smoothed and stage-specific abnormality performance is integrated to obtain event attention, making the prediction more universal and robust for new patients.

[0077] Preferably, in one embodiment of the present invention, considering that the greater the fluctuation range of blood pressure abnormality during the hypotension impact phase, it indicates that the blood pressure abnormality fluctuates drastically, reflecting that the blood pressure has experienced drastic fluctuations and the degree of abnormality is higher; at the same time, the greater the maximum blood pressure abnormality within the phase, it indicates that the operating mode has caused more severe hypotension consequences and the degree of abnormality is higher.

[0078] Based on this, in each stage of hypotension, the abnormal impact span is obtained according to the fluctuation range of blood pressure abnormality in each stage of hypotension; the abnormal impact span of each stage of hypotension and the maximum blood pressure abnormality are combined to obtain the stage-specific abnormality performance.

[0079] As an example, the range of blood pressure abnormality in each hypotension stage is linearly normalized in the corresponding data dimension, and the normalized result is used as the abnormality span. The abnormality span of each hypotension stage is multiplied and fused, and the product of the abnormality span of each hypotension stage and the maximum blood pressure abnormality is used as the stage-specific abnormality of the corresponding hypotension stage, reflecting the distribution of blood pressure abnormality in a hypotension stage.

[0080] Furthermore, in order to determine the degree of abnormal blood pressure manifestations caused by a certain type of operational mode, and to smooth out individual and physician operational differences, the overall characteristics of the degree of abnormal manifestations in all stages of each type of hypotension are used to obtain the event attention level, which is more representative of the typical and common risk level of this operational mode.

[0081] As an example, the overall characteristics of the data are presented as an average value, reflecting the distribution of blood pressure abnormality in a certain type of hypotension impact phase. The average value of all phases of abnormality in each type of hypotension impact phase is taken as the event attention level for that type of hypotension impact phase.

[0082] In other embodiments of the present invention, the implementer may also obtain the event attention level by weighted summation of the average, median and mode of the staged abnormal performance, for example by weighted summation with weights of 0.5, 0.2 and 0.3.

[0083] The real-time early warning module 103 is used to construct a real-time operation event set by constructing all surgical operations within the preset historical neighborhood of the current patient's latest surgical operation; in the hypotension impact stage that best matches the real-time operation event set, the operation-blood pressure linkage impact degree is obtained by combining the similarity characteristics between the real-time operation event set and the surgical operation of each hypotension impact stage, combined with the stage abnormality degree and event attention degree, and real-time early warning is carried out with the help of a neural network model.

[0084] After determining the event attention levels for different categories of hypotension impact stages, we can analyze the real-time operations of the current surgical patient. By using the most matching historical classification, we can assess the impact of the current operation on blood pressure. Therefore, we construct a real-time operation event set by collecting all surgical operations within the preset historical neighborhood of the current patient's latest surgical operation. This avoids focusing only on the current operation or the most recent surgical operation, and captures the synergistic and cumulative effects between operations, providing basic data for subsequent pattern matching.

[0085] As an example, the preset time length of the historical neighborhood is 10 minutes. During the operation, whenever the doctor performs a new operational event (surgical operation), the operation is traced back 10 minutes from the starting point of the latest surgical operation, and the relevant surgical operation is added to the real-time operation event set. The real-time operation event set is then updated once, and the real-time operation event set contains the latest surgical operation.

[0086] It should be noted that in other embodiments of the present invention, the implementer may adjust the time length of the preset historical neighborhood.

[0087] After obtaining the current real-time operation event set, we can match the most similar historical hypotension impact stage, thereby using historical data analysis to predict the impact of the current operation on blood pressure, providing more reference for the neural network model. Considering that different historical hypotension impact stages have different reference values ​​for the real-time operation event set, and that the stage abnormality performance and event attention reflect the impact of the surgical operation on blood pressure from both local and overall perspectives.

[0088] Therefore, in the hypotension impact stage that best matches the real-time operation event set, based on the similarity characteristics between the real-time operation event set and the surgical operation in each hypotension impact stage, combined with the stage abnormality performance and event attention, the operation-blood pressure linkage impact is obtained, and real-time early warning is provided with the help of a neural network model.

[0089] Preferably, in one embodiment of the present invention, the most matching low blood pressure impact stage in the real-time operation event set is obtained according to the nearest neighbor algorithm.

[0090] As an example, the average and maximum values ​​of the Jaccard similarity coefficients between the real-time operation event set and the surgical operations of each hypotension impact stage are obtained. The sum of the average and maximum values ​​of the Jaccard similarity coefficients corresponding to each hypotension impact stage is used as the matching coefficient between the real-time operation event set and each hypotension impact stage. The hypotension impact stage with the highest matching coefficient is selected.

[0091] It should be noted that the Jaccard similarity coefficient is a well-known technique and will not be elaborated upon further.

[0092] Preferably, in one embodiment of the present invention, please refer to Figure 3 The flowchart illustrates a method for obtaining the operational-blood pressure linkage effect according to an embodiment of the present invention, specifically including:

[0093] Step S301: In the hypotension impact stage that best matches the real-time operation event set, obtain the comparison reference weight based on the similarity characteristics between the real-time operation event set and the surgical operation of each hypotension impact stage.

[0094] Considering that different historical periods of low blood pressure have different reference values ​​for the real-time operation event set, the historical periods with higher similarity of surgical operations have higher reference weights. Therefore, the comparative reference weights are obtained first.

[0095] As an example, the Jaccard correlation coefficient between the surgical procedures in the real-time operation event set and the surgical procedures in each hypotension stage is used as a comparative reference weight. The Jaccard correlation coefficient is used to characterize the similarity features between the real-time operation event set and the surgical procedures in each hypotension stage.

[0096] Step S302: Integrate and compare the reference weights and the corresponding stage abnormality performance of the hypotension impact stage to obtain the real-time operation event set and the corresponding local linkage impact degree of each hypotension impact stage.

[0097] Considering that the higher the reference weight of the hypotension impact stage, the higher the degree of stage-specific abnormality, it indicates that the current real-time operation event set is more likely to reproduce the high-risk state represented by that historical stage. Therefore, this is used to obtain the degree of local linkage impact.

[0098] As an example, the product of the comparative reference weight and the corresponding stage of abnormal performance of the hypotension impact stage is used as the local linkage impact degree between the real-time operation event set and each hypotension impact stage. From a local perspective, this characterizes the stage of abnormal performance of a single historical hypotension impact stage and predicts the linkage impact degree of the current real-time operation event set on blood pressure.

[0099] It should be noted that the real-time operational event set is consistent with the comparative analysis process for each stage of hypotension impact; only one example is described here.

[0100] Step S303: Based on the overall characteristics of the local linkage influence degree and combined with the event attention degree, obtain the operation-blood pressure linkage influence degree.

[0101] After step S302 compares the real-time operation event set with the most matching hypotension impact stage one by one, the linkage impact of the current real-time operation event set on blood pressure is evaluated from the perspective of a single historical hypotension impact stage. Further prediction and evaluation from an overall perspective are needed, so the overall characteristics of local linkage impact are utilized. At the same time, considering that event attention represents the degree of attention required for the historical hypotension impact stage and reflects the overall impact of the current most matching hypotension impact stage on blood pressure, the operation-blood pressure linkage impact is also obtained by combining event attention.

[0102] As an example, after linearly normalizing the average value of the local linkage impact on the corresponding data dimension, the normalization result is multiplied by the event attention, and the product is used as the operation-blood pressure linkage impact of the real-time operation event set.

[0103] In this way, the feature of the operation-blood pressure linkage influence of the current patient's real-time operation event set can be added to the traditional neural network to supplement the input to the neural network, and real-time early warning can be achieved with the help of the neural network model.

[0104] Since the training and use of neural network models are well-known techniques, only a brief description is provided here:

[0105] Collect a large dataset of cardiac surgery patient medical records from different periods (e.g., more than 1,000). For these datasets, extract the relevant physiological data of the patients (i.e., blood pressure, heart rate, body temperature, etc.) and the operation-blood pressure linkage impact of the corresponding real-time operation event set of the patients. Then, perform manual evaluation to determine the corresponding patient hypotension risk warning level, which is used as the output of the neural network.

[0106] After processing all datasets, the training set and validation set are divided in a 7:3 ratio. The neural network is trained using the training set samples with the cross-entropy function as the loss function. Gradient descent is used to train until the loss function converges. The robustness of the training results is verified using the validation set, and the trained neural network is obtained.

[0107] This allows us to leverage current neural networks to provide timely warnings of potential low blood pressure risks after doctors perform new procedures, thus preventing secondary harm to patients caused by inaccurate intraoperative low blood pressure warnings.

[0108] In another embodiment of the present invention, a preset warning threshold, such as 0.75, can be set. When the real-time operation-blood pressure linkage influence exceeds the preset warning threshold, a warning is issued. In other embodiments of the present invention, the implementer can also set a graded warning method, which will not be elaborated further.

[0109] In summary, to address the technical problem of existing intraoperative hypotension warning methods failing to predict the impact of intraoperative procedures in a timely manner, resulting in a certain lag in warnings, this invention provides a monitoring information analysis system for surgical procedures. This invention first acquires blood pressure monitoring data and obtains the analysis interval for each intraoperative event; further, it analyzes the hypotensive manifestation of blood pressure in each event to obtain the blood pressure abnormality degree; further, based on the distribution of blood pressure abnormality degree in historical patient data for each type of hypotension impact stage, it obtains the stage-specific abnormality degree and integrates it to obtain the event attention level; further, it acquires the latest real-time operation event set for the current patient, and in the most matching hypotension impact stage, based on the similarity characteristics between the real-time operation event set and the surgical procedures in each hypotension impact stage, combined with the stage-specific abnormality degree and event attention level, it obtains the operation-blood pressure linkage impact degree, and uses a neural network model for real-time warnings. This solution dynamically quantifies the potential impact of surgical procedures by analyzing the linkage between the real-time operation event set and the most matching hypotension impact stage, and inputs this information into a neural network to achieve proactive warnings, solving the problem of warning lag caused by ignoring the impact of procedures and individual differences in existing systems, thus ensuring patient safety.

[0110] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0111] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A monitoring information analysis system for surgical procedures, characterized in that, The system includes: Monitoring module: acquires monitoring data for three preset blood pressure values; annotates surgical procedures at various stages during the operation and divides the analysis interval for each event; Analysis module: Based on the low-pressure manifestation of the three blood pressure types within the analysis interval of each event, obtain the blood pressure abnormality degree; in the historical patient data, divide the low blood pressure impact stage according to the fluctuation of the blood pressure abnormality degree; classify the low blood pressure impact stage according to the similarity characteristics of surgical operations between different low blood pressure impact stages; obtain the stage-specific abnormality performance degree of each low blood pressure impact stage according to the distribution of the blood pressure abnormality degree of each category of low blood pressure impact stage, and merge them to obtain the event attention degree; Real-time early warning module: Constructs a real-time operation event set by including all surgical operations within a preset historical neighborhood of the patient's latest surgical operation; In the hypotension impact stage that best matches the real-time operation event set, Based on the similarity characteristics between the real-time operation event set and the surgical operations of each hypotension impact stage, combined with the stage-specific abnormality performance and the event attention, obtains the operation-blood pressure linkage impact degree, and inputs the operation-blood pressure linkage impact degree and the patient's real-time physiological monitoring data into a neural network model for real-time early warning; The method for classifying the hypotension impact stages includes: obtaining the clustering distance between two hypotension impact stages based on the intersection-union ratio of surgical operations between any two hypotension impact stages; performing clustering based on all the clustering distances; and having each cluster correspond to a classification of a hypotension impact stage. The method for obtaining the event attention includes: in each type of hypotension impact stage, obtaining the abnormal impact span based on the fluctuation range of the blood pressure abnormality in each hypotension impact stage; and fusing the abnormal impact span of each hypotension impact stage with the maximum blood pressure abnormality to obtain the stage-specific abnormal performance. The average of all the stage-specific abnormal manifestations in each of the hypotension-affecting stages is taken as the event attention level; The method for obtaining the operation-blood pressure linkage influence includes: in the hypotension influence stage that best matches the real-time operation event set, obtaining a comparison reference weight based on the similarity characteristics between the real-time operation event set and the surgical operation of each hypotension influence stage; By integrating the comparative reference weights and the corresponding stage-specific abnormality of the hypotension impact stage, the local linkage impact degree of the real-time operation event set and each hypotension impact stage is obtained. Based on the overall characteristics of the local linkage influence, and combined with the event attention, the operation-blood pressure linkage influence is obtained.

2. The monitoring information analysis system for surgical procedures according to claim 1, characterized in that, The method for obtaining the blood pressure abnormality includes: Within the analysis interval of each event, each type of blood pressure is taken as the target blood pressure, and data where the target blood pressure is lower than the preset diastolic reference threshold are marked as abnormal data. Based on the time proportion of the abnormal data and the area between the data curve of the abnormal data and the reference line of the preset diastolic reference threshold, the diastolic abnormality of the target blood pressure is obtained. By integrating the low-pressure abnormality manifestations of the three blood pressure data, the blood pressure abnormality of the corresponding event is obtained.

3. The monitoring information analysis system for surgical procedures according to claim 2, characterized in that, The method for obtaining the blood pressure abnormality level of the corresponding event by fusing the low blood pressure abnormality levels of the three blood pressure values ​​includes: For each event, the normalized result of the low-pressure abnormality of the mean arterial pressure is used as the first weight, the difference between constant 1 and the first weight is used as the second weight, the low-pressure abnormality of the systolic pressure is weighted based on the first weight, the low-pressure abnormality of the diastolic pressure is weighted based on the second weight, and the weighted sum is used as the blood pressure abnormality of the corresponding event.

4. The monitoring information analysis system for surgical procedures according to claim 1, characterized in that, The method for obtaining the comparison reference weights includes: The Jaccard correlation coefficient between the surgical operations of the real-time operation event set and the surgical operations of each of the hypotension-affected stages is used as a comparative reference weight.

5. A monitoring information analysis system for surgical procedures according to claim 1, characterized in that, The most matching class of hypotension-affecting stages in the real-time operation event set is obtained based on the nearest neighbor algorithm.

6. A monitoring information analysis system for surgical procedures according to claim 1, characterized in that, The method for obtaining the hypotension effect phase includes: In the data of each historical patient, the time domain is divided into different stages of hypotension influence by using the maximum value of the blood pressure abnormality in the time domain as the dividing point.

7. A monitoring information analysis system for surgical procedures according to claim 1, characterized in that, The methods for obtaining the analysis interval include: The time interval between the start time of each surgical procedure and the start time of the next surgical procedure is used as the analysis interval for each event.

Images