Preoperative risk assessment method for cardiology

Abstract

This invention relates to the field of physiological signal prediction technology, specifically a method for preoperative risk assessment in cardiology, comprising the following steps: calculating baseline offset by segmenting time-series data using a sliding window; generating anomaly markers by dynamically warping and aligning the fluctuation rates of parameters; triggering risk signals by comparing standard deviation with amplitude thresholds; mapping risk levels using K-means clustering of multidimensional data; detecting abnormal fluctuations using isolated forest and cross-validating with electrocardiogram and myocardial enzyme profiles; and outputting a preoperative risk assessment conclusion. In this invention, the sliding window algorithm eliminates individual differences and interference; dynamic time warping aligns the fluctuation rates of multiple parameters to address time variability; comparing standard deviation with amplitude thresholds establishes a quantitative assessment standard to avoid the limitations of a single threshold; K-means clustering combined with Euclidean distance achieves objective grading; and the isolated forest algorithm, combined with cross-validation of the ST segment of electrocardiogram and myocardial enzyme profiles, forms a multimodal assessment system, creating a complete closed loop that improves the sensitivity and specificity of risk assessment.

Description

Technical Field

[0001] This invention relates to the field of physiological signal prediction technology, and in particular to a method for preoperative risk assessment in cardiology. Background Technology

[0002] The field of physiological signal prediction technology encompasses technical methods for monitoring and predicting human health status using various physiological parameters and signals. These physiological signals include electrocardiograms, blood pressure, pulse, and respiratory rate, collected through sensors or monitoring devices. The core of physiological signal prediction technology lies in the real-time analysis and processing of the collected data to extract health information with early warning capabilities. This technology is primarily applied in medical monitoring, disease prevention, and health management, using scientific analysis to predict potential health risks and support medical decision-making. Key aspects of physiological signal prediction technology include signal acquisition, signal processing, data analysis, and model prediction, aiming to provide conditions for early intervention through accurate data prediction.

[0003] Among them, preoperative risk assessment methods in cardiology refer to a technical approach that assesses preoperative risk by analyzing patients' physiological signals. This topic mainly addresses the technical issues in preoperative risk assessment, specifically covering the use of patients' physiological signal data, such as electrocardiograms, blood pressure, and pulse, combined with clinical background information, to conduct multi-dimensional analysis of their preoperative state. This method establishes mathematical models, combining patients' raw health data and real-time monitoring data, to analyze and predict the health risks that patients may experience before surgery. This approach provides medical personnel with objective evidence for preoperative risk assessment, enhancing the scientific rigor of surgical safety management.

[0004] Current technologies for analyzing physiological signals largely rely on single-parameter static threshold judgments, lacking multi-parameter dynamic coordination mechanisms. This makes it difficult to effectively capture abnormal correlations between differentiated physiological parameters. For example, fluctuations in ECG and blood pressure data differ over time, and traditional methods, without time alignment processing, are prone to misjudgments. Existing risk assessment models often use fixed or empirical thresholds, failing to adapt to individual physiological differences and easily leading to missed detections or false alarms when patients' baseline levels fluctuate significantly. Risk level classification relies heavily on expert experience or simple segmentation rules, lacking data-driven quantitative classification criteria, resulting in overly general grading results that are difficult to support refined clinical decision-making. Existing methods primarily rely on single-modality data for abnormal signal verification, failing to fully integrate clinical indicators such as ECG ST segment and myocardial enzyme profiles, resulting in data silos and reducing the clinical interpretability of prediction results. For example, abnormal myocardial enzyme profiles without correlation analysis with real-time physiological signals delay early diagnosis of myocardial ischemia. These shortcomings lead to imbalances in sensitivity and specificity, and delayed risk assessment in practical applications of existing technologies. Summary of the Invention

[0005] To address the shortcomings of existing technologies that rely heavily on single-parameter static threshold judgments for physiological signal analysis, lacking a multi-parameter dynamic coordination mechanism, and thus failing to effectively capture abnormal correlations between differentiated physiological parameters (e.g., the temporal differences in fluctuations between ECG and blood pressure data, which traditional methods lack time alignment processing for, easily leading to misjudgments), and to address the issue that current methods often employ fixed or empirical thresholds, failing to adapt to individual physiological differences and prone to missed detections or false alarms when patients' baseline levels fluctuate significantly, and to address the issue that risk level classification often relies on expert experience or simple segmentation rules, lacking data-driven quantitative classification criteria, resulting in overly generalized classifications that are difficult to support refined clinical decision-making, and to address the issue that existing methods primarily rely on single-modality data for abnormal signal verification, failing to fully integrate clinical indicators such as ECG ST segment and myocardial enzyme profiles, resulting in data silos and reduced clinical interpretability of prediction results (e.g., abnormal myocardial enzyme profiles without correlation analysis with real-time physiological signals can delay early diagnosis of myocardial ischemia), these deficiencies lead to technical problems such as an imbalance between sensitivity and specificity and delayed risk assessment in practical applications. Therefore, this invention provides a preoperative risk assessment method for cardiology. The technical solution is as follows:

[0006] On the one hand, it provides a method for preoperative risk assessment in cardiology, which includes:

[0007] S1: Obtain the patient's preoperative physiological parameters such as heart rate, blood pressure and respiratory rate through medical monitoring equipment, segment the time series data using the sliding window algorithm, calculate the median within the window as the baseline, and perform offset processing on the physiological parameter data and the baseline to obtain the fluctuation offset.

[0008] S2: Based on the fluctuation offset, perform time alignment processing based on the dynamic time warping algorithm, calculate the fluctuation rate of each parameter, determine whether there is a trend abnormal change, and obtain a trend abnormality mark.

[0009] S3: Call the trend anomaly marker, extract the corresponding fluctuation offset to evaluate the fluctuation amplitude, compare it with the amplitude threshold, and generate a risk trigger signal when two or more fluctuation amplitudes exceed the amplitude threshold;

[0010] S4: Call the risk trigger signal, extract multidimensional physiological parameter data for the corresponding time period, classify them using the K-means clustering algorithm, set three cluster centers, calculate the risk level by using Euclidean distance as the standard, and output the risk level label;

[0011] S5: Call the risk level label, use the isolated forest algorithm to detect the degree of abnormal fluctuation offset of the time-series record data corresponding to the risk level label, combine the ECG ST segment and myocardial enzyme spectrum detection data to perform clinical indication cross-validation, and output the preoperative risk assessment conclusion.

[0012] As a further embodiment of the present invention, the sliding window duration of the sliding window algorithm is 5 minutes, and the window sliding step size is 1 minute.

[0013] Before clustering, the input data needs to be Z-score standardized. The standard deviation screening threshold is set in combination with actual data and clinical statistics, and low, medium and high risk level labels are output.

[0014] The trend anomaly is defined based on a fluctuation rate threshold set at twice the standard deviation of the original data mean. When the rates of two or more parameters exceed the threshold, it is determined to be a trend anomaly.

[0015] The isolated forest algorithm combines the fluctuation trajectory obtained by the dynamic time warping algorithm with abnormal time periods to perform spatiotemporal analysis.

[0016] The fluctuation offset specifically refers to the baseline offset of heart rate, baseline offset of blood pressure, and baseline offset of respiratory rate. The trend anomaly marker includes a yes / no Boolean marker. The risk trigger signal includes a yes / no Boolean value identifier. The risk level label specifically refers to the low-risk threshold range, the medium-risk transition range, and the high-risk warning range. The risk assessment conclusion includes the set of data points assessed as high-risk before surgery.

[0017] As a further aspect of the present invention, the specific steps of S1 include:

[0018] S101: Acquire physiological data recorded by monitoring equipment, including heart rate, blood pressure, and respiratory rate; sort the data based on timestamps; segment the time-series data using a sliding window; calculate the median within the window; and generate a window baseline median sequence.

[0019] S102: Based on the window reference median sequence, call the time series data, perform offset calculation point by point, obtain the offset between the data and the window reference value, and generate a physiological parameter offset sequence;

[0020] S103: Based on the offset data of heart rate, blood pressure and respiratory rate in the physiological parameter offset sequence, extract the large offset value, small offset value and mean offset value, mark the corresponding monitoring time in time order, calculate the offset fluctuation range of each physiological parameter, and obtain the fluctuation offset.

[0021] The offset fluctuation range was compared and analyzed with the original data from 72 hours before surgery to obtain the baseline reference value of the movement line.

[0022] As a further aspect of the present invention, the offset is calculated using the following formula:

[0023] ;

[0024] in, This represents the offset at the nth time point. This represents the physiological parameter measurement value at the nth time point. This represents the median value of the w-th window in the baseline median sequence of the sliding window. This represents the arithmetic mean of the physiological parameter measurements within the w-th sliding window. This represents the physiological parameter measurement value within the w-th sliding window. The value is a dynamic regulatory factor determined based on the patient's age and gender, ranging from 0.8 to 1.2.

[0025] As a further aspect of the present invention, the specific steps of S2 include:

[0026] S201: Based on the fluctuation offset, the dynamic time warping algorithm is used to perform time series alignment processing on the two-cycle parameter values, identify the synchronous offset trajectory of the parameters at different time points, and generate aligned time series feature values.

[0027] S202: Based on the aligned time series feature values, calculate the change range of the parameters in a continuous time period, extract the change slope of the parameter values ​​in the time period, compare the change slope of each parameter with its baseline slope in the aligned time series, calculate the slope deviation of the parameters in the time period, and obtain the slope offset difference value.

[0028] The baseline slope was obtained by training the rate of change of parameters under normal conditions 72 hours before surgery.

[0029] S203: Based on the slope offset difference value, set a trend change identification threshold, determine whether the difference value of the parameter exceeds the trend change identification threshold, and uniformly mark and classify the parameters within the time period that continuously exceeds the threshold to generate trend anomaly markers.

[0030] The threshold for identifying trend changes is set with reference to clinical standards and based on expert experience.

[0031] As a further aspect of the present invention, the specific steps of S3 include:

[0032] S301: Call the abnormal fluctuation marker dataset, extract the fluctuation offset values ​​of the parameters within the marked time period, obtain the offset sequence of the parameters at the time node, fit the offset vector to calculate the offset variance, judge the fluctuation stability based on the offset variance, and obtain the fluctuation offset difference data.

[0033] S302: Based on the fluctuation offset difference data, calculate the standard deviation of the parameter offset value within the marked time period, call the amplitude threshold set by the parameter, calculate the parameter whose standard deviation value exceeds the corresponding amplitude threshold, and obtain the offset amplitude threshold excess data;

[0034] S303: Call the offset amplitude threshold out-of-tolerance data, set the lower limit value for the judgment of the number of out-of-tolerance, compare the number of marked parameters within the time period according to the lower limit value, and generate a risk trigger signal based on the joint judgment of parameter offset amplitude and number of out-of-tolerance.

[0035] The amplitude threshold is set to twice the standard deviation of the offset within the original sliding window of the parameter.

[0036] As a further aspect of the present invention, the specific steps of S4 include:

[0037] S401: Call the risk trigger signal to obtain multidimensional physiological parameter data for the required time period. Based on the clustering requirements, set three cluster centers, calculate the Euclidean distance value from each data point to the three cluster centers, and use the shortest distance method to classify the data points into the corresponding categories to obtain the Euclidean distance mapping result.

[0038] S402: Based on the Euclidean distance mapping result, extract the distance values ​​corresponding to the data points within the cluster category, use the maximum and minimum distance difference to divide the risk gradient within and between groups, set the distance range boundary to classify the three cluster center values ​​into three risk levels respectively, and obtain risk level classification data.

[0039] S403: Based on the risk level classification data, map and match the risk level correspondence between each cluster label and the cluster center, and replace the label to which each data point is classified to obtain the risk level label.

[0040] As a further aspect of the present invention, the Euclidean distance value is calculated using the following formula:

[0041] ;

[0042] in, This represents the Euclidean distance from the i-th data point to the j-th cluster center; it is a dimensionless parameter. This represents the physiological parameter value of the i-th data point in the k-th dimension. Let represent the coordinate value of the j-th cluster center in the k-th dimension, which is a dimensionless parameter. This represents the standard deviation of the k-th dimension of data. Represents the total number of data dimensions; it is a dimensionless parameter. This represents the dynamic adjustment coefficient, which is a dimensionless parameter.

[0043] As a further aspect of the present invention, the specific steps of S5 include:

[0044] S501: Call the risk level label, detect the degree of offset fluctuation anomaly based on the isolated forest algorithm, extract the fluctuation offset data of the abnormal parameters, filter the fluctuation anomaly amount and mark the parameters in combination with the detection results, and obtain the offset fluctuation anomaly amount data.

[0045] S502: Based on the aforementioned offset fluctuation abnormality data, according to the ECG ST segment and myocardial enzyme spectrum detection data, extract the physiological indicator data of each abnormal parameter, calculate the cross-validation index based on the detection data, mark the cross-validation parameters that meet the abnormal conditions, and obtain abnormal cross-validation data.

[0046] The electrocardiogram ST segment and myocardial enzyme spectrum detection data were standardized using Z-score with unified dimensions.

[0047] S503: Based on the abnormal cross-validation data, perform preoperative risk analysis on the cross-validation parameters, and combine the common characteristics of fluctuation deviation and abnormal physiological indicators to obtain the preoperative risk assessment conclusion.

[0048] As a further aspect of the present invention, the cross-validation metric is calculated using the following formula:

[0049] ;

[0050] in, The cross-validation data represents the q-th parameter. This represents the Z-score standardized value of the ECG ST segment detection data corresponding to the q-th parameter. This represents the Z-score standardized value of the myocardial enzyme spectrum detection data corresponding to the q-th parameter. This represents the abnormal value of the offset fluctuation of the q-th parameter. The arithmetic mean of the abnormal fluctuations in the representative parameter offset. This represents the standardized ST segment value of the electrocardiogram at time point k for the q-th parameter. This represents the standardized value of the myocardial enzyme spectrum at time point k for the q-th parameter. This represents the total number of time points.

[0051] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0052] By segmenting time-series data using a sliding window algorithm and calculating baseline offset, the influence of individual differences and transient interference on physiological parameters is effectively eliminated, enhancing data stability. A dynamic time warping algorithm is employed to align multi-parameter fluctuation rates, addressing the misjudgment of anomalies caused by inconsistent time dimensions in traditional methods and improving the accuracy of multi-parameter collaborative analysis. Based on a comparison mechanism of standard deviation and amplitude threshold, a quantitative assessment standard for multi-dimensional fluctuation amplitude is established, avoiding the limitations of single-threshold judgment and enhancing the robustness of anomaly detection. K-means clustering is used to classify multi-dimensional physiological parameter data, and Euclidean distance is used to map risk levels, achieving data-driven objective grading and reducing reliance on subjective experience. The isolated forest algorithm is combined with cross-validation of abnormal fluctuation offsets and ECG ST segment and myocardial enzyme spectrum data, overcoming the bottleneck of traditional single-dimensional analysis and forming a multi-modal data fusion assessment system, improving the clinical applicability of preoperative risk prediction. These techniques form a complete closed loop from data preprocessing, dynamic alignment, quantitative assessment to multi-modal validation, improving the sensitivity and specificity of risk assessment. Attached Figure Description

[0053] Figure 1 This is a schematic diagram of the workflow of the present invention. Detailed Implementation

[0054] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0055] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0056] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0057] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0058] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0059] Please see Figure 1 This invention provides a method for preoperative risk assessment in cardiology, and the process of this method may include the following steps:

[0060] S1: Obtain the patient's preoperative physiological parameters such as heart rate, blood pressure and respiratory rate through medical monitoring equipment, segment the time series data using the sliding window algorithm, calculate the median within the window as the baseline, and perform offset processing on the physiological parameter data and the baseline to obtain the fluctuation offset.

[0061] S2: Based on the fluctuation offset, perform time alignment processing based on the dynamic time warping algorithm, calculate the fluctuation rate of each parameter, determine whether there is a trend-related abnormal change, and obtain a trend-related abnormality marker.

[0062] S3: Call the trend anomaly marker, extract the corresponding fluctuation offset to evaluate the fluctuation amplitude, and compare it with the amplitude threshold. When two or more fluctuation amplitudes exceed the amplitude threshold, a risk trigger signal is generated.

[0063] S4: Invoke the risk trigger signal, extract the multidimensional physiological parameter data of the corresponding time period, classify them using the K-means clustering algorithm, set three cluster centers, calculate the risk level by using Euclidean distance as the standard, and output the risk level label;

[0064] S5: Call the risk level label, use the isolated forest algorithm to detect the abnormal degree of fluctuation offset of the time-series record data corresponding to the risk level label, combine the ECG ST segment and myocardial enzyme spectrum detection data to perform clinical indication cross-validation, and output the preoperative risk assessment conclusion.

[0065] The fluctuation offset specifically includes the baseline offset of heart rate, blood pressure, and respiratory rate. The trend abnormality marker includes a yes / no Boolean marker. The risk trigger signal includes a yes / no Boolean value marker. The risk level label specifically refers to the low-risk threshold range, the medium-risk transition range, and the high-risk warning range. The risk assessment conclusion includes the set of data points that are assessed as high-risk before surgery.

[0066] Specifically, the steps of S1 are as follows:

[0067] S101: Acquire physiological data recorded by monitoring equipment, including heart rate, blood pressure, and respiratory rate; sort the data based on timestamps; segment the time-series data using a sliding window; calculate the median within the window; and generate a window baseline median sequence.

[0068] To obtain physiological data recorded by monitoring equipment, it is first necessary to identify the types of parameters recorded by the equipment, including heart rate (unit: beats / minute), blood pressure (unit: mmHg, composed of systolic and diastolic blood pressure), and respiratory rate (unit: breaths / minute). The monitoring data is recorded in timestamp format and stored in a local database or server system. Taking a certain ECG monitoring device as an example, the heart rates recorded between 9:00 and 10:00 on May 18, 2025, were 78, 82, 85, 83, 88, 90, 92, 85, 80, and 76 beats / minute, corresponding to timestamps recorded at 60-second intervals. The program reads the recorded data for each minute within this time period and sorts it according to chronological order. The sorting process involves rearranging the data items using an index. After sorting, the timestamps are matched one-to-one with their corresponding heart rate, blood pressure, and respiratory rate to form an ordered triplet array. Then, a sliding window technique is used to segment the time-series data. Taking a window length of 5 minutes and a sliding step of 1 minute as an example, the heart rate values ​​in the first window are 78, 82, 85, 83, and 88. Using Python's slicing logic, the oldest data is removed and a new record is added after each slide. For example, the second window has values ​​of 82, 85, 83, 88, and 90, and so on, forming multiple sliding windows. The median of heart rate, blood pressure, and respiratory rate data is calculated sequentially for each window. Taking heart rate as an example, the median of the window sequence {78, 82, 85, 83, 88} is 85, which is recorded as the baseline value for the first window. The median calculation involves sorting the values ​​within the window from smallest to largest and taking the middle value. If the number of values ​​is even, the average of the two middle values ​​is taken. For example, the sorting of the window {85, 83, 88, 90, 92} is {83, 85, 88, 90, 92}, with a median of 88. Arranging the medians of the windows sequentially forms the window baseline median sequence, as shown in the following example:

[0069] Table 1 Monitoring Data and Median Sequence Table

[0070] Time window start time Heart rate sequence (beats / minute) Median heart rate (beats / minute) Systolic blood pressure sequence (mmHg) Median systolic blood pressure (mmHg) Respiratory rate sequence (mmHg) Median respiratory rate (mmHg) 09:00 78，82，85，83，88 85 120，125，130，128，122 125 18，20，19，20，19 19 09:01 82，85，83，88，90 85 125，130，128，122，118 125 20，19，20，19，18 19 09:02 85，83，88，90，92 88 130，128，122，118，115 122 19，20，19，18，17 19

[0071] As shown in Table 1, the median calculation within the sliding window can be used to smooth physiological signal fluctuations and serve as a benchmark for subsequent offset calculations. Once the median sequence within the window is formed, it can be further used for data offset processing. The obtained median value will be used as a fixed benchmark in the difference calculation with the original data points.

[0072] In refining the specific execution actions of the phrases "sorting," "splitting," and "calculating," the sorting operation first involves reading the timestamp field of the recorded data and sorting it in ascending order. The corresponding index array is then rearranged according to time sequence, allowing subsequent sliding window operations to continuously extract data along the timeline. The splitting operation involves traversing the entire array, setting an initial pointer to the first data point, and defining the window size. Step length Each segment starts from the current position. arrive Extract data and store it as a list, then slide it backwards in a loop. Each unit generates the next window, repeating this process until the end of the data. The final median calculation involves taking all data within the window, calling a sorting function to obtain the data item corresponding to the middle index value. If the data volume is even, the median is obtained by averaging the two middle values. For example, with the heart rate window {78, 82, 85, 83, 88}, after sorting, it becomes {78, 82, 83, 85, 88}, with a median of 83, which is assigned as the window's baseline value. The process of obtaining the term "median" can be further quantified: with five records {82, 85, 87, 90, 92}, after sorting, it becomes {82, 85, 87, 90, 92}, with the middle item being the third item, 87; while for six records {82, 85, 87, 90, 92, 95}, after sorting, it becomes {82, 85, 87, 90, 92, 95}, with a median of (87+90) / 2=88.5, which is assigned to the window. If a window of data contains extreme outliers (such as a sudden increase in heart rate to 40 beats per minute) during the segmentation and median calculation process, the original value should be retained and the data should proceed to the subsequent offset evaluation stage.

[0073] S102: Based on the window baseline median sequence, call time series data, perform offset calculations point by point, obtain the offset between the data and the window baseline value, and generate a physiological parameter offset sequence;

[0074] Offset processing based on the window baseline median sequence first requires iterating through the original record values ​​at each time point. And identify the window to which it belongs. Extract the median corresponding to the window. mean with standard deviation The offset is calculated using the following formula:

[0075] ;

[0076] in, This represents the offset at the nth time point. This represents the physiological parameter measurement value at the nth time point. This represents the median value of the w-th window in the baseline median sequence of the sliding window. This represents the arithmetic mean of the physiological parameter measurements within the w-th sliding window. This represents the physiological parameter measurement value within the w-th sliding window. The value is a dynamic regulatory factor determined based on the patient's age and gender, ranging from 0.8 to 1.2.

[0077] For each The offset calculation is performed as follows: Taking heart rate as an example, the recorded value at time 09:00 is 78 beats / minute, and the window is 09:00–09:04. The median value within the window is... mean Standard deviation The calculation is as follows:

[0078] ;

[0079] The patient was set to be male, 55 years old, and the pathology data was reviewed and set accordingly. Then we can obtain:

[0080] ;

[0081] Similarly, perform the above operations for each time point and obtain the offset to ensure that the processing logic is executed separately for heart rate, blood pressure, and respiratory rate. Calculate the offsets for systolic and diastolic blood pressure separately. For example, if the systolic blood pressure is 120 mmHg at 09:00, the corresponding window sequence is {120, 125, 130, 128, 122}, and calculate the median. mean Standard deviation The calculation is as follows:

[0082] ;

[0083] The offset is calculated as follows:

[0084] ;

[0085] Offsets are used to characterize the dynamic fluctuations of multiple physiological parameters. After obtaining the offset data, it needs to be organized in a structured manner according to timestamps. For example, at 09:00, the heart rate offset is 2.663, the systolic blood pressure offset is 2.634, the diastolic blood pressure offset is 1.918, and the respiratory rate offset is 0.789. The above data is stored in vector form. The corresponding time point is 09:00, forming a physiological parameter offset sequence; further, each offset value is classified and divided, and a judgment interval is set based on the absolute value of the offset: if the heart rate offset value is... Defined as a slight offset, if A value greater than 5 indicates a moderate deviation, while a value greater than 5 indicates a significant deviation. The interval setting here references clinical statistical results from large-sample heart rate fluctuation studies, and the corresponding blood pressure parameter setting interval is a slight deviation. Medium offset: (3 ， 8] Significant deviation: greater than 8 breaths / minute; slight deviation in respiratory rate: Medium offset: (2 ， 5] Significant offsets (greater than 5 times / minute) are identified, and this offset data is used to generate a classifiable vector dataset. It's important to note that when the recording frequency increases to once every 30 seconds, but the window remains 5 minutes with a 1-minute step, the number of data points within each window increases. Therefore, it's still necessary to locate the median, mean, and standard deviation of each data point within its nearest time window, recalculate the corresponding offset, and improve the recording time precision to the second level for high-precision monitoring and processing of individual physiological data. The output offset value sequence should be a set of vectors with timestamps, for example: 09:00 → [2.663, 2.634, 1.918, 0.789], 09:01 → [1.322, 2.541, 1.110, 0.552], achieving time-based offset calculation and processing of all physiological parameters.

[0086] S103: Based on the offset data of heart rate, blood pressure and respiratory rate in the physiological parameter offset sequence, extract the large offset value, small offset value and mean offset value, mark the corresponding monitoring time in time order, calculate the offset fluctuation range of each physiological parameter, and obtain the fluctuation offset.

[0087] The range of offset fluctuations was compared with the original data from 72 hours before surgery to obtain the reference value of the movement baseline.

[0088] For each physiological parameter, the generated offset sequence needs to have its maximum, minimum, and mean values ​​extracted, and its corresponding time points marked. The processing flow involves slicing the vector data according to the parameter dimension. For example, for heart rate, the heart rate offset terms at each time point are extracted to form an offset value array. Execute the maximum value function to select the maximum offset value. Call the minimum value function to get the minimum offset value. Simultaneously, it traces back to the point in time where the event occurred using the corresponding index, and then executes the mean function: The overall trend mean is obtained. For example, in the heart rate offset sequence {−3.2, −1.1, +0.9, +2.3, −2.5}, the maximum value is +2.3, the minimum value is −3.2, corresponding to the times 09:03 and 09:00, and the average value is... The results showed that the overall heart rate shift was biased towards the negative direction, and the variability shift was calculated as follows:

[0089] ,

[0090] Systolic blood pressure, diastolic blood pressure, and respiratory rate are processed separately in this manner, and three types of extreme values ​​and means are extracted. After the fluctuation offset is formed, the maximum, minimum, and mean positions and corresponding values ​​at time points are recorded using structured vectors and tables. After obtaining the fluctuation offset of the parameters in the current monitoring period, the database is called to obtain the patient's complete monitoring records for the 72 hours before surgery. Data collected by the same monitoring device is matched, and the same sliding window length and sliding step size as the current data are used for segmentation and median extraction. The preoperative baseline median sequence is reconstructed, and the same calculation process as S102 is performed to obtain the preoperative physiological parameter offset value sequence. Then, the preoperative fluctuation offset is extracted to form the preoperative baseline data. The current monitoring fluctuation value is directly subtracted from the preoperative fluctuation baseline value. For example, if the current respiratory rate fluctuation is 7 breaths / minute and the preoperative baseline fluctuation is 3 breaths / minute, the difference is 1 / 3. The baseline reference value assigned to the parameters was +4, indicating that the degree of fluctuation was increased compared to the preoperative level. The difference values ​​can be organized into a vector. These correspond to physiological indicators such as heart rate, systolic blood pressure, diastolic blood pressure, and respiratory rate, respectively, for subsequent patient dynamic trend modeling and analysis.

[0091] Specifically, the steps of S2 are as follows:

[0092] S201: Based on the fluctuation offset, the dynamic time warping algorithm is used to perform time series alignment processing on the two-period parameter values, identify the synchronous offset trajectory of the parameters at different time points, and generate aligned time series feature values.

[0093] Based on the fluctuation offset between the current cycle and the preoperative cycle, parameter time series alignment is required. During this process, the time series sampling values ​​of a specific monitoring parameter (such as heart rate) within the current cycle are first set as the master sequence. For example, heart rate values ​​at five time points from 09:00 to 09:04 are collected: 78, 81, 85, 79, and 82 beats / minute. Simultaneously, a sequence with the same parameter collected within the 72 hours preoperatively is selected as control data, such as the preoperative sequence: 75, 76, 77, 76, and 78 beats / minute. Then, the two sets of sequences are arranged into time-parameter value key-value pairs according to their chronological order to ensure time point consistency. Sampling is performed at minute intervals. If the number of sampling points in a sequence is insufficient, linear interpolation is used to fill the gaps. In this example, the two sequences have the same length, so the filling operation is skipped. Next, to achieve accurate numerical matching, the alignment deviation of the two sequences needs to be compared. The distance metric is set to absolute difference, i.e., at any time point... Below, current value Compared with control value The offset is A fifth-order distance matrix is ​​constructed based on this. The matching paths are traversed, and the best mapping group is selected based on the principle of minimizing the cumulative distance of the paths. In this example, since the time points are completely aligned, a one-to-one correspondence is directly formed without the need to rearrange the paths. The complete aligned sequence mapping pairs are as follows: 09:00 corresponds to 75 before surgery, 09:01 corresponds to 76, 09:02 corresponds to 77, 09:03 corresponds to 76, and 09:04 corresponds to 78. The parameter data in the two time periods are synchronized and combined through the alignment structure to obtain the aligned time series table shown in Table 2 for subsequent processing.

[0094] S202: Based on the feature values ​​of the aligned time series, calculate the magnitude of the parameter change in a continuous time period, extract the slope of the parameter value change in the time period, compare the slope of each parameter change with its baseline slope in the aligned time series, calculate the slope deviation of the parameter in the time period, and obtain the slope offset difference value.

[0095] The baseline slope was obtained by training the rate of change of parameters under normal conditions 72 hours before surgery.

[0096] After completing the parameter alignment operation, it is necessary to further evaluate the slope of the parameter changes between consecutive time points to reveal whether there are any abnormal deviations in the trend. During the execution process, the numerical difference between two adjacent time points in the current sequence is extracted first, that is, the slope of the change in heart rate from 78 to 81 between time points 09:00 and 09:01 is... Similarly, 09:01–09:02 is 09:02–09:03 09:03–09:04 The slope of the current sequence is The same calculations were performed on the preoperative sequence to obtain its slope sequence. That is Then, the difference between the two sets of slope values ​​is calculated to obtain the offset slope sequence. Construct a complete offset difference array Subsequently, a threshold judgment standard was set. Referring to actual monitoring experience and expert settings, the trend identification slope threshold was set to 2 times / minute. The absolute value of each difference was processed and compared with the threshold. The following judgments were made sequentially: at 09:01, the slope offset was 2, equal to the threshold, and not considered exceeding the threshold; at 09:02, the offset was 3, exceeding the threshold; at 09:03, the offset was -5, with an absolute value of 5, exceeding the threshold; at 09:04, the offset was 1, not exceeding the threshold. This resulted in a set of trend offset Boolean sequences. The specific results are shown in Table 2.

[0097] S203: Based on the slope offset difference value, set a trend change identification threshold, determine whether the difference value of the parameter exceeds the trend change identification threshold, and uniformly mark and classify the parameters within the time period that continuously exceeds the threshold to generate trend anomaly markers.

[0098] The threshold for identifying trend changes is set with reference to clinical standards and based on expert experience;

[0099] After obtaining the trend offset Boolean judgment result sequence, it is necessary to analyze whether there are consecutive "yes" states in the sequence, thereby identifying time periods with continuous offset trends. Specifically, this involves traversing the slope offset judgment Boolean sequence and checking the current sequence. In the analysis, two consecutive time points, 09:02 and 09:03, were identified as "yes," confirming a trend shift within the interval. Anomaly marking is required, and the process is as follows: First, record the start and end times of the consecutive "yes" interval (09:02 to 09:03 in this case). Then, extract the original slope data and baseline slope data for the time period. The current slope changes are 4 and -6, the baseline slopes are 1 and -1, and the slope shift values ​​are 3 and -5. The time period and corresponding parameter values ​​are then entered into a structured data table and marked as "trend anomaly" for subsequent analysis. In practice, this process requires simultaneous analysis of multiple parameters. After marking, the results are updated in conjunction with the original parameter sequence. The trend anomaly set can be expressed as: This time period constitutes the only trend-shifting interval in this embodiment.

[0100] Table 2 Alignment Sequence and Trend Offset Data Table

[0101] Time point Current heart rate (beats / minute) Baseline heart rate (beats / minute) Current slope of change (times / minute) Baseline change slope (times / minute) Slope offset difference value Does it exceed the threshold? Trend Anomaly Marking 09:00 78 75 09:01 81 76 3 1 2 no 09:02 85 77 4 1 3 yes yes 09:03 79 76 -6 -1 -5 yes yes 09:04 82 78 3 2 1 no

[0102] As shown in Table 2, by identifying trend deviations and calculating slope differences, it is clear that there is an abnormal trend in the current period from 09:02 to 09:03.

[0103] Specifically, the steps of S3 are as follows:

[0104] S301: Call the abnormal fluctuation label dataset, extract the fluctuation offset values ​​of the parameters within the label time period, obtain the offset sequence of the parameters at the time node, fit the offset vector to calculate the offset variance, judge the fluctuation stability based on the offset variance, and obtain the fluctuation offset difference data.

[0105] After calling the trend anomaly labeled dataset, offset analysis needs to be performed on the heart rate parameters within the labeled time period from 09:02 to 09:03. First, the heart rate values ​​for this time period are extracted from the current period and the baseline period, respectively. The current period values ​​are 85 and 79, and the baseline periods are 77 and 76. The offset of the time point is then calculated accordingly; for example, the offset for 09:02 is... The offset at 09:03 is , forming an offset sequence Next, the offset sequence needs to be fitted with an offset vector. The specific operation is as follows: the offset data is used as a sample vector for centering, that is, the mean is subtracted from each of the offset data. , thus obtaining the centered vector Then, the vector is squared to obtain... The average of these values ​​is then taken to obtain the offset variance. The calculation process is as follows: Therefore, the variance of the offset for this time period is 6.25. If this value is used as the fluctuation offset difference data, it needs to be compared with the mean difference in the original sample data. For example, if the original mean variance is 3.0, then the difference value for this time period is significantly higher than the mean. Based on the difference magnitude, it is determined that the parameter fluctuation magnitude in the current time period does not belong to the normal range. This result can be further used as input for offset magnitude assessment to execute the next step of analysis. The specific offset extraction operation needs to be traversed sequentially according to the abnormal trend time periods. If there are multiple abnormal segments, an independent offset sequence is constructed for each segment and its variance value is calculated. For example, the newly added segment from 09:10 to 09:11 has heart rate values ​​of 90 and 85, corresponding to baselines of 82 and 81, with offsets of 8 and 4, and a mean of 6. The offset variance is calculated as follows: Thus, it can be seen that different segments have different degrees of offset fluctuation. The difference between the above offset variance and the mean of the whole sample is used as a measure of the degree of fluctuation offset difference.

[0106] S302: Based on the fluctuation offset difference data, calculate the standard deviation of the parameter offset value within the marked time period, call the amplitude threshold set by the parameter, calculate the parameter whose standard deviation value exceeds the corresponding amplitude threshold, and obtain the offset amplitude threshold excess data;

[0107] Based on the offset variance obtained in the previous stage as the fluctuation offset difference data, the standard deviation is calculated for the offset sequence within each abnormal time period. Specifically, for the offset sequence from 09:02 to 09:03... To process this, first calculate the mean, which is 5.5, and then calculate the standard deviation:

[0108] ;

[0109] The standard deviation is then compared with a preset offset amplitude threshold, which is defined as twice the standard deviation of the offset within the original sliding window. If the total offset obtained in the current monitoring period using a 5-minute sliding window is... Its standard deviation is calculated as The mean is 4 and the variance is: ;

[0110] The offset magnitude threshold is Since the standard deviation from 09:02 to 09:03 is 2.5, which is lower than 2.82, it is not considered out of tolerance. If the offset of another segment is... The mean is 7 and the standard deviation is . If the value is greater than the threshold of 2.82, it is marked as out-of-tolerance data. This process is repeated for each trend anomaly segment, comparing its standard deviation and recording the results to generate out-of-tolerance data for the offset magnitude threshold. The recorded data can be stored in a labeled array, such as... The second paragraph is out of tolerance, while the first and third paragraphs are normal.

[0111] S303: Call the offset amplitude threshold out-of-tolerance data, set the lower limit value for the judgment of the number of out-of-tolerance, compare the number of marked parameters within the time period based on the lower limit value, and generate a risk trigger signal based on the joint judgment of parameter offset amplitude and number of out-of-tolerance.

[0112] The amplitude threshold is set to twice the standard deviation of the offset within the original sliding window of the parameter;

[0113] After calling the offset amplitude threshold deviation data, it is necessary to further set the lower limit value for the number of deviations and perform joint judgment. In the current example, there are a total of three abnormal trend time periods. The lower limit value is set to 1 item. That is, if more than one parameter exceeds the amplitude deviation in the same abnormal trend segment, the risk is judged to have reached the trigger standard. In actual operation, it is necessary to perform statistics on each abnormal trend segment. For example, the offset standard deviation corresponding to the segment 09:02 to 09:03 is 2.5, which does not exceed the 2.82 threshold and is marked as "No". Another segment, 09:10 to 09:11, has 3, which is marked as "Yes". After statistics, it is found that there is 1 deviation in this segment, which is equal to the set lower limit value of 1, which meets the trigger condition and generates a risk trigger signal. If the third segment data has an offset of A standard deviation of 0 does not constitute an out-of-tolerance condition and is not recorded as a trigger. The entire judgment operation requires traversing the trend anomaly segments, calculating the standard deviation and counting out-of-tolerance for the parameters within each segment, and comparing it with the lower limit value after the statistics are completed. For example, if the lower limit value is set to 1, and the number of out-of-tolerance parameters is... If this occurs, a "risk trigger" signal is generated. This signal can be identified by a Boolean value and used as an input parameter for subsequent responses, forming a unified risk response control mechanism.

[0114] Table 3. Statistics of Offset in Abnormal Trend Sections

[0115] Time period Current parameter value Reference parameter values Offset sequence mean variance Standard deviation Amplitude threshold Is it out of tolerance? Risk Trigger Marker 09:02–09:03 85，79 77，76 8，3 5.5 6.25 2.5 2.82 no no 09:10–09:11 90，85 82，81 8，4 6 8 2.83 2.82 yes yes 09:15–09:16 85，85 80，80 5，5 5 0 0 2.82 no no

[0116] As shown in Table 3, some abnormal trend segments have exceeded the threshold limit in terms of amplitude standard deviation, and are marked as out of tolerance and trigger the corresponding risk signal.

[0117] Specifically, the steps of S4 are as follows:

[0118] S401: Call the risk trigger signal, obtain multidimensional physiological parameter data for the required time period, set three cluster centers based on clustering requirements, calculate the Euclidean distance value from each data point to the three cluster centers, and use the shortest distance method to classify the data points into the corresponding categories to obtain the Euclidean distance mapping results.

[0119] The system invokes a risk trigger signal, indicating that an alarm has been received regarding abnormal preoperative signs in the patient, such as rapid changes in heart rate or arrhythmia within 24 hours prior to surgery. At this point, the system automatically retrieves the patient's continuous 24-hour preoperative monitoring data, covering multiple physiological parameters including heart rate, blood pressure, respiratory rate, and blood oxygen saturation. Each dimension acquires 24 consecutive data points per hour, forming a 24×m multidimensional physiological parameter data matrix. Assuming m=4, representing four different types of preoperative physiological signal parameters, the system first extracts the 24-hour data sequence for each dimension and calculates its standard deviation. For example, the extracted data sequence for the heart rate dimension is [76, 79, 82, ..., 73], and the calculated standard deviation is... By analogy, the standard deviation of heart rate can be obtained as follows: Based on the clustering analysis requirements, the system sets three cluster centers to represent low-risk, medium-risk, and high-risk preoperative states, respectively. It constructs three sets of cluster center values ​​based on the average preoperative characteristic values ​​of previous medical records, and sets them as follows: , , Each set of center points represents the average standard model of a class of preoperative conditions. For any preoperative data point... For example, the physiological parameter data of a patient at the 10th hour is Then, it is necessary to calculate the improved Euclidean distance between it and the three cluster centers, using the formula:

[0120] ;

[0121] in, This represents the Euclidean distance from the i-th data point to the j-th cluster center; it is a dimensionless parameter. This represents the physiological parameter value of the i-th data point in the k-th dimension. Let represent the coordinate value of the j-th cluster center in the k-th dimension, which is a dimensionless parameter. This represents the standard deviation of the k-th dimension of data. Represents the total number of data dimensions; it is a dimensionless parameter. This represents the dynamic adjustment coefficient, which is a dimensionless parameter.

[0122] During execution, the difference is first extracted and standardized according to the dimension, for example... Continuing to process the dimensions in this manner, the normalized differences are as follows: Squaring the four terms and then summing them, we get , take the root of Introducing dynamic adjustment factors With dimension Calculate the adjustment coefficient To arrive at the cluster center The distance is And so on, calculate to and The distance results are 1.924 and 3.366 respectively. The category 1 corresponding to the minimum value of 1.410 is taken as the cluster label of the data point. To ensure completeness, the system performs the above calculation process on the 24-hour data points in sequence, and records the category to which the shortest distance of each data point belongs as shown below.

[0123] Table 4. Euclidean Distance Mapping Table

[0124] Data point number Distance to center 1 (dimensionless) Distance to center 2 (dimensionless) Distance to the center 3 (dimensionless) Minimum distance category <![CDATA[X 10 ]]> 1.410 1.924 3.366 Category 1 <![CDATA[X 11 ]]> 1.685 1.392 2.914 Category 2 <![CDATA[X 12 ]]> 2.154 2.471 1.863 Category 3

[0125] As shown in Table 4, the clustering distance results of three monitoring data points of cardiology patients during a certain period before surgery are listed. The category corresponding to the minimum value is the Euclidean distance mapping classification result of that point. See Table 4. This result is the basic data for subsequent risk level classification.

[0126] S402: Based on the Euclidean distance mapping results, extract the distance values ​​corresponding to data points within the cluster category, use the maximum and minimum distance difference to divide the risk gradient within and between groups, set the distance range boundary to classify the three cluster center values ​​into three risk levels, and obtain risk level classification data.

[0127] Based on the completed Euclidean distance mapping results, the system needs to extract the minimum Euclidean distance value of data points in each cluster category within their respective category. For example, in a preoperative assessment scenario, category 1 contains data points... , The minimum distance values ​​corresponding to these values ​​are 1.410, 1.295, etc., respectively, forming the set of distance values ​​for this category. The system extracts the minimum Euclidean distance sets for categories 2 and 3 respectively. Then, for each cluster's minimum distance set, it calculates the difference between the maximum and minimum values ​​within each group, forming the minimum-maximum distance difference for that category. For example, if the maximum value in category 1 is 1.689 and the minimum is 1.295, then the distance difference for that category is... Similarly, the difference in distance between category 2 and category 3 is calculated to obtain... , The three types of interval differences are sorted in ascending order to obtain... This indicates that the risk gradient within a group ranges from low to high. The system then introduces a unified risk level classification rule between groups, performing linear standardization on the minimum distance value intervals within each category. This divides the minimum-maximum interval of each category into three risk level intervals. For example, the distance value interval [1.295, 1.689] for category 1 is divided into: low risk [1.295, 1.426], medium risk [1.426, 1.557], and high risk [1.557, 1.689]. Following this process, the interval boundaries are calculated using the formula:

[0128] ;

[0129] in, Representing the The upper boundary value of each interval. Representing the The minimum Euclidean distance value among cluster categories. Representing the The maximum value of the Euclidean distance among the cluster categories. The index represents the interval division. Two dividing lines are calculated in the order of 1 to 2 to divide the entire numerical interval into three level intervals at equal intervals.

[0130] Taking category 2 as an example, if its minimum value is 1.392 and its maximum value is 2.003, then the boundary... , The intervals [1.392, 1.595), [1.595, 1.797), and [1.797, 2.003] are marked as low, medium, and high risk levels, respectively. Based on this, the distance value of each data point is judged to determine its risk level label in its respective cluster category, thereby completing the risk level classification of the three types of centroids and their contained data. This classification process does not involve cross-class comparisons, but only relies on the risk gradient interval defined by the intra-group distance distribution characteristics to achieve the risk level classification data output of preoperative data.

[0131] S403: Based on the risk level classification data, map and match the risk level correspondence between each cluster label and the cluster center, and replace the corresponding label of each data point to obtain the risk level label.

[0132] Based on the aforementioned risk level classification data, label replacement is performed on the data points corresponding to each cluster label. During this process, a matching mapping table between cluster categories and their risk level labels needs to be established. For example, the three risk levels for category 1 are L1-1 (low), L1-2 (medium), and L1-3 (high), corresponding to the intervals within each category that fall into based on the minimum Euclidean distance. Each data point's original label is "Category 1," but if its distance value within category 1 falls into the L1-2 interval, its label needs to be updated to "medium risk." Similarly, for category 2 and category 3 data, mapping is performed to the three sub-levels L2-1L2-3 and L3-1L3-3 for "medium risk" and "high risk," respectively. By replacing the original category labels through this mapping, the risk level label corresponding to each data point is formed. For example, data points... Originally classified as category 1, with a distance value of 1.410, falling within the interval [1.295, 1.426), corresponding to a low-risk level, labeled "low-risk," while the data point... Originally belonging to category 2, with a distance value of 1.392, falling within the interval [1.392, 1.595), it is also mapped to "low risk". It should be noted here that even if the risk level labels are the same, but their categories are different, the label update only reflects the update at the risk level level and does not affect the subsequent grouping distinction. After the system completes the risk label update operation of the data points, it obtains the risk level label set of the preoperative data, which is used for subsequent risk intervention strategy formulation and preoperative decision reference.

[0133] Specifically, the steps in S5 are as follows:

[0134] S501: Call the risk level label, detect the degree of deviation fluctuation anomaly based on the isolated forest algorithm, extract the fluctuation deviation data of the abnormal parameters, filter the fluctuation anomaly amount and mark the parameters in combination with the detection results, and obtain the deviation fluctuation anomaly amount data.

[0135] After calling the risk level label, it is necessary to locate the data points marked as medium-high risk in each cluster category. The system first extracts the sampling data of each parameter in the data points in multiple detection cycles, and extracts the time-series record data of monitoring parameters such as heart rate, blood pressure, and blood oxygen saturation in turn. Then, it performs outlier detection processing between samples for each parameter. During the execution process, the system first calls the set of observed values ​​of the parameter in different detection cycles. For example, the heart rate record value of a certain high-risk data point in five cycles from T1 to T5 is [78, 80, 76, 122, 79]. The system then performs maximum, minimum and median value extraction operations on the array to obtain 122, 76 and 79 respectively. Then the system calculates the set of offset values ​​of the parameter in the sample, that is, the absolute value of the difference between each value and the median value [1]. [1, 3, 43, 0], determine whether the offset value exceeds the preset offset reference value. For example, if the offset reference value is set to 20, the system will only keep the record with an offset value of 43 based on this judgment and mark the corresponding period T4 as an abnormal fluctuation point. Then, the same process is performed on the parameters. A corresponding identifier is formed for the abnormal period corresponding to each abnormal parameter. At the same time, the absolute value of the offset and the positive and negative identifier of the offset direction of each parameter are recorded as the data structure of the extraction result. Furthermore, the system accumulates and counts the number of all parameters detected as abnormal in each data point and sets the threshold for the number of abnormal parameters to 2. When a data point has two or more parameters marked as abnormal, the system assigns an offset fluctuation abnormal label to the data point and outputs the set of data points that meet the condition, which is the offset fluctuation abnormal amount data set.

[0136] S502: Based on the abnormal offset fluctuation data, according to the ECG ST segment and myocardial enzyme spectrum detection data, extract the physiological indicator data of each abnormal parameter, calculate the cross-validation index based on the detection data, mark the cross-validation parameters that meet the abnormal conditions, and obtain abnormal cross-validation data.

[0137] Electrocardiogram ST segment and myocardial enzyme spectrum data were standardized using Z-score with unified dimensions.

[0138] Based on the abnormal deviation fluctuation data and the standardized ECG ST segment and myocardial enzyme spectrum data, each abnormal parameter needs to be extracted item by item during the process. The detected values ​​at a specified time point are then subjected to Z-score standardization to obtain... and For example, a parameter such as "troponin I" is obtained after standardization. , Offset fluctuation anomaly The average value of the abnormal parameter offset fluctuation Calculate the cross-validation metric using the formula:

[0139] ;

[0140] in, The cross-validation data represents the q-th parameter. This represents the Z-score standardized value of the ECG ST segment detection data corresponding to the q-th parameter. This represents the Z-score standardized value of the myocardial enzyme spectrum detection data corresponding to the q-th parameter. This represents the abnormal value of the offset fluctuation of the q-th parameter. The arithmetic mean of the abnormal fluctuations in the representative parameter offset. This represents the standardized ST segment value of the electrocardiogram at time point k for the q-th parameter. This represents the standardized value of the myocardial enzyme spectrum at time point k for the q-th parameter. This represents the total number of time points.

[0141] The first part is calculated as follows: Then, parameters were extracted at time points. The detected values ​​are used to calculate the standardized sum of squared sequence differences, assuming... The detection data at each time point is shown below:

[0142] Table 5 Standardized values ​​of electrocardiogram and myocardial enzyme data

[0143] Time point ST segment standardized value Enzyme spectrum normalization values T1 1.80 1.00 T2 1.95 1.05 T3 1.60 0.98 T4 1.70 1.12

[0144] As shown in Table 5, the parameters fluctuated significantly in the ST segment and enzyme spectrum data at different time points, and the calculated results showed that... A value higher than the threshold indicates that the parameter has a large deviation under different detection methods and time dimensions, which is a risk signal that requires close attention.

[0145] Then their squared differences are as follows: , , , Add the above values ​​together and divide by the number of time points. Then perform the square root operation, that is The two parts were combined and calculated to obtain This value is the cross-validation data value for the troponin I parameter. If the risk assessment threshold is set to 1.2, and this threshold is based on the statistical regularity of the original sample data, an optimal cutoff line is determined by analyzing the distribution of cross-validation values ​​between normal and abnormal samples to ensure the stability and discriminative power of the classification. Therefore, this parameter is determined to be an abnormal cross-validation parameter and is marked in the system to form abnormal cross-validation data.

[0146] S503: Based on the abnormal cross-validation data, perform preoperative risk analysis on the cross-validation parameters, and combine the common characteristics of fluctuation deviation and abnormal physiological indicators to obtain the preoperative risk assessment conclusion.

[0147] Based on the abnormal cross-validation data, the system needs to read the labeling information of each cross-validation parameter and analyze its corresponding original detection parameter data. First, the system identifies each detection index marked as an abnormal cross-validation parameter and extracts its corresponding ECG ST segment standardized data and myocardial enzyme spectrum standardized data. For example, if a parameter is known to be marked as abnormal at a certain data point, and its ECG ST segment Z-score is 1.85 and its myocardial enzyme spectrum Z-score is 0.92, then the system extracts these two values ​​and calculates the difference between them. This is a structural difference between the corresponding detection methods. Next, the corresponding offset fluctuation anomaly value is called. Assuming the value is 0.58, then retrieve all abnormal parameters. arithmetic mean of values Let its value be 0.64, and then we can complete the first part of the calculation. The Z-score data for this parameter is further extracted at multiple time points. Assuming the Z-scores for electrocardiogram and enzyme spectrum at four time points are [1.80, 1.95, 1.60, 1.70] and [1.00, 1.05, 0.98, 1.12] respectively, the system calculates the square of the difference between the two at each time point and then averages them. Then perform square root calculation. The cross-validation data value is obtained by adding the two parts together. The system then sets the cross-validation threshold to 1.2. This threshold is determined based on a large-scale statistical analysis and ROC (Receiving Controller Operating Characteristic) curve optimization process using the preoperative risk assessment model on the original sample data. A judgment operation is then performed; if... If the parameter is abnormal, it is added to the preoperative risk parameter tag set. After the abnormal item calculation is completed, the system counts the number of abnormal parameters of the data points in turn. If the number is greater than the set risk judgment value of 2, the data point is marked as a preoperative high-risk object. Otherwise, the label is not updated, and the set of data points assessed as preoperative high-risk is output to form the preoperative risk assessment conclusion.

[0148] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention 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 the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for preoperative risk assessment in cardiology, characterized in that, Includes the following steps: S1: Obtain the patient's preoperative physiological parameters such as heart rate, blood pressure and respiratory rate through medical monitoring equipment, segment the time series data using the sliding window algorithm, calculate the median within the window as the baseline, and perform offset processing on the physiological parameter data and the baseline to obtain the fluctuation offset. S2: Based on the fluctuation offset, perform time alignment processing based on the dynamic time warping algorithm, calculate the fluctuation rate of each parameter, determine whether there is a trend abnormal change, and obtain a trend abnormality mark. S3: Call the trend anomaly marker, extract the corresponding fluctuation offset to evaluate the fluctuation amplitude, compare it with the amplitude threshold, and generate a risk trigger signal when two or more fluctuation amplitudes exceed the amplitude threshold; S4: Call the risk trigger signal, extract multidimensional physiological parameter data for the corresponding time period, classify them using the K-means clustering algorithm, set three cluster centers, use Euclidean distance as the standard to calculate the cluster center value to map the risk level, and output the risk level label; S5: Call the risk level label, use the isolated forest algorithm to detect the degree of abnormal fluctuation offset of the time-series record data corresponding to the risk level label, combine the ECG ST segment and myocardial enzyme spectrum detection data to perform clinical indication cross-validation, and output the preoperative risk assessment conclusion.

2. The method for preoperative risk assessment in cardiology according to claim 1, characterized in that, The sliding window algorithm has a sliding window duration of 5 minutes and a window sliding step size of 1 minute. Before clustering, the input data needs to be Z-score standardized. The standard deviation screening threshold is set by combining real-time data and clinical statistics, and low, medium and high risk level labels are output. The trend anomaly is defined based on a fluctuation rate threshold set at twice the standard deviation of the original data mean. When the rates of two or more parameters exceed the threshold, it is determined to be a trend anomaly. The isolated forest algorithm combines the fluctuation trajectory obtained by the dynamic time warping algorithm with abnormal time periods to perform spatiotemporal analysis. The fluctuation offset specifically refers to the baseline offset of heart rate, baseline offset of blood pressure, and baseline offset of respiratory rate. The trend anomaly marker includes a yes / no Boolean marker. The risk trigger signal includes a yes / no Boolean value identifier. The risk level label specifically refers to the low-risk threshold range, the medium-risk transition range, and the high-risk warning range. The risk assessment conclusion includes the set of data points assessed as high-risk before surgery.

3. The method for preoperative risk assessment in cardiology according to claim 1, characterized in that, The specific steps of S1 include: S101: Acquire physiological data recorded by monitoring equipment, including heart rate, blood pressure, and respiratory rate; sort the data based on timestamps; segment the time-series data using a sliding window; calculate the median within the window; and generate a window baseline median sequence. S102: Based on the window reference median sequence, call the time series data, perform offset calculation point by point, obtain the offset between the data and the window reference value, and generate a physiological parameter offset sequence; S103: Based on the offset data of heart rate, blood pressure and respiratory rate in the physiological parameter offset sequence, extract the large offset value, small offset value and mean offset value, mark the corresponding monitoring time in time order, calculate the offset fluctuation range of each physiological parameter, and obtain the fluctuation offset. The offset fluctuation range was compared and analyzed with the original data from 72 hours before surgery to obtain the baseline reference value of the movement line.

4. The method for preoperative risk assessment in cardiology according to claim 3, characterized in that, The offset is calculated using the following formula: ； in, This represents the offset at the nth time point. This represents the physiological parameter measurement value at the nth time point. This represents the median value of the w-th window in the baseline median sequence of the sliding window. This represents the arithmetic mean of the physiological parameter measurements within the w-th sliding window. This represents the physiological parameter measurement value within the w-th sliding window. It is a dynamic regulatory factor determined based on the patient's age and gender.

5. The method for preoperative risk assessment in cardiology according to claim 3, characterized in that, The specific steps of S2 include: S201: Based on the fluctuation offset, the dynamic time warping algorithm is used to perform time series alignment processing on the two-cycle parameter values, identify the synchronous offset trajectory of the parameters at different time points, and generate aligned time series feature values. S202: Based on the aligned time series feature values, calculate the change range of the parameters in a continuous time period, extract the change slope of the parameter values ​​in the time period, compare the change slope of each parameter with its baseline slope in the aligned time series, calculate the slope deviation of the parameters in the time period, and obtain the slope offset difference value. The baseline slope was obtained by training the rate of change of parameters under normal conditions 72 hours before surgery. S203: Based on the slope offset difference value, set a trend change identification threshold, determine whether the difference value of the parameter exceeds the trend change identification threshold, and uniformly mark and classify the parameters within the time period that continuously exceeds the threshold to generate trend anomaly markers. The threshold for identifying trend changes is set with reference to clinical standards and based on expert experience.

6. The method for preoperative risk assessment in cardiology according to claim 5, characterized in that, The specific steps of S3 include: S301: Call the trend anomaly marker, extract the fluctuation offset value of the parameter within the marked time period, obtain the offset sequence of the parameter at the time node, fit the offset vector to calculate the offset variance, judge the fluctuation stability based on the offset variance, and obtain the fluctuation offset difference data. S302: Based on the fluctuation offset difference data, calculate the standard deviation of the parameter offset value within the marked time period, call the amplitude threshold set by the parameter, calculate the parameter whose standard deviation value exceeds the corresponding amplitude threshold, and obtain the offset amplitude threshold excess data; S303: Call the offset amplitude threshold out-of-tolerance data, set the lower limit value for the judgment of the number of out-of-tolerance, compare the number of marked parameters within the time period according to the lower limit value, and generate a risk trigger signal based on the joint judgment of parameter offset amplitude and number of out-of-tolerance. The amplitude threshold is set to twice the standard deviation of the offset within the original sliding window of the parameter.

7. The method for preoperative risk assessment in cardiology according to claim 6, characterized in that, The specific steps of S4 include: S401: Call the risk trigger signal to obtain multidimensional physiological parameter data for the required time period. Based on the clustering requirements, set three cluster centers, calculate the Euclidean distance value from each data point to the three cluster centers, and use the shortest distance method to classify the data points into the corresponding categories to obtain the Euclidean distance mapping result. S402: Based on the Euclidean distance mapping result, extract the distance values ​​corresponding to the data points within the cluster category, use the maximum and minimum distance difference to divide the risk gradient within and between groups, set the distance range boundary to classify the three cluster center values ​​into three risk levels respectively, and obtain risk level classification data. S403: Based on the risk level classification data, map and match the risk level correspondence between each cluster label and the cluster center, and replace the label to which each data point is classified to obtain the risk level label.

8. The method for preoperative risk assessment in cardiology according to claim 7, characterized in that, The Euclidean distance value is calculated using the following formula: ； in, This represents the Euclidean distance from the i-th data point to the j-th cluster center, and is a dimensionless parameter. This represents the physiological parameter value of the i-th data point in the k-th dimension. Let represent the coordinate value of the j-th cluster center in the k-th dimension, which is a dimensionless parameter. This represents the standard deviation of the k-th dimension of data. Represents the total number of data dimensions; it is a dimensionless parameter. This represents the dynamic adjustment coefficient, which is a dimensionless parameter.

9. The method for preoperative risk assessment in cardiology according to claim 7, characterized in that, The specific steps of S5 include: S501: Call the risk level label, detect the degree of offset fluctuation anomaly based on the isolated forest algorithm, extract the fluctuation offset data of the abnormal parameters, filter the fluctuation anomaly amount and mark the parameters in combination with the detection results, and obtain the offset fluctuation anomaly amount data. S502: Based on the aforementioned offset fluctuation abnormality data, according to the ECG ST segment and myocardial enzyme spectrum detection data, extract the physiological indicator data of each abnormal parameter, calculate the cross-validation index based on the detection data, mark the cross-validation parameters that meet the abnormal conditions, and obtain abnormal cross-validation data. The electrocardiogram ST segment and myocardial enzyme spectrum detection data were standardized using Z-score with unified dimensions. S503: Based on the abnormal cross-validation data, perform preoperative risk analysis on the cross-validation parameters, and combine the common characteristics of fluctuation deviation and abnormal physiological indicators to obtain the preoperative risk assessment conclusion.

10. The method for preoperative risk assessment in cardiology according to claim 9, characterized in that, The cross-validation metric is calculated using the following formula: ； in, The cross-validation data represents the q-th parameter. This represents the Z-score standardized value of the ECG ST segment detection data corresponding to the q-th parameter. This represents the Z-score standardized value of the myocardial enzyme spectrum detection data corresponding to the q-th parameter. This represents the abnormal value of the offset fluctuation of the q-th parameter. The arithmetic mean of the abnormal fluctuations in the representative parameter offset. This represents the standardized ST segment value of the electrocardiogram at time point k for the q-th parameter. This represents the standardized value of the myocardial enzyme spectrum at time point k for the q-th parameter. This represents the total number of time points.

Images