A liver cirrhosis complication risk intelligent evaluation and early warning method

Abstract

The present application relates to the technical field of monitoring and evaluation, in particular to a liver cirrhosis complication risk intelligent evaluation and early warning method. The method generates a sequence of hemodynamic records by feature recognition of electrocardiogram signals and arterial blood pressure signals, constructs a ventricle-artery event matching cost matrix and solves the optimal matching sequence. On the basis of the optimal matching sequence, the low blood pressure event risk probability is analyzed and early warning is carried out by analyzing the structural disorder index and functional disorder index. The present application quantifies the structural disorder and functional disorder indexes of ventricle-artery coupling relationship, analyzes the low blood pressure event risk probability and early warning, realizes the precise quantitative early warning of hemodynamic abnormalities, effectively improves the overall efficiency of monitoring, and provides reliable data support for clinical precise intervention.

Description

Technical Field

[0001] This invention relates to the field of monitoring and assessment technology, specifically to a method for intelligent assessment and early warning of the risk of complications of cirrhosis. Background Technology

[0002] Cirrhosis is the end stage of many chronic liver diseases. Complications caused by portal hypertension are the main cause of poor prognosis for patients. Transjugular intrahepatic portosystemic shunt (TIPS) is a key interventional procedure for treating this disease. However, in the early postoperative period, the patient's systemic hemodynamic environment will undergo drastic remodeling, and the circulatory system needs to adapt to this change, which can easily induce serious complications such as acute hypotension.

[0003] Traditional monitoring techniques primarily rely on threshold alarms for macroscopic indicators such as mean arterial pressure. However, these indicators typically reflect the final outcome of compensatory dysfunction and exhibit a certain lag. Before macroscopic indicators deteriorate, the stable causal relationship between ventricular ejection and arterial response is often disrupted. Current technologies cannot capture and quantify this dynamic transition from "orderly coupling" to "loss of conduction consistency," resulting in lagging and one-sided risk monitoring that fails to meet the urgent clinical need for early and precise intervention in high-risk patients. Summary of the Invention

[0004] To address the shortcomings of existing technologies, such as the lack of early disorder analysis to identify a stable causal relationship between ventricular ejection and arterial system response, resulting in monitoring lag and incomplete information, this invention aims to provide an intelligent assessment and early warning method for the risk of cirrhosis complications. The specific technical solution adopted is as follows: This invention provides a method for intelligent assessment and early warning of the risk of complications from cirrhosis, the method comprising: Acquire electrocardiogram and arterial blood pressure signals synchronously in time sequence, identify cardiac cycles and assign global pulsation indexes, extract key features of arterial blood pressure waveforms, and generate a beat-by-beat hemodynamic recording sequence; The recent pulsation analysis window is extracted from the beat-by-beat hemodynamic recording sequence. The diastolic waveform of each cardiac cycle in the analysis window is decoupled from the physiological model, and the arterial system response time constant is extracted. The ventricular-arterial event matching cost matrix containing time cost, morphological cost and response characteristic cost is constructed, and the optimal event matching sequence is obtained by solving it. Based on the difference between the optimal event matching sequence and the natural time series, structural disorder indicators are determined; by analyzing the correlation stability between pulse pressure and response time constant under the optimal event matching sequence, functional disorder indicators are determined. By aggregating time-series structural disorder indicators and functional disorder indicators, the risk probability of hypotension events is obtained through a pre-trained risk assessment model; and early warning is provided based on the risk probability of hypotension events.

[0005] Furthermore, the beat-by-beat hemodynamic recording sequence is composed of single-beat hemodynamic recordings arranged in the order of the global pulsation index; each single-beat hemodynamic recording encapsulates data related to a single cardiac cycle, including a global pulsation index, a set of key timestamps, pulse pressure, and diastolic waveform data; the set of key timestamps includes the R-wave peak timestamp, the systolic blood pressure peak timestamp, and the descending isthmus notch timestamp, and the pulse pressure is the difference between the systolic blood pressure value and the diastolic blood pressure value.

[0006] Furthermore, the method for obtaining the optimal event matching sequence includes: A ventricular-arterial event matching cost matrix is ​​constructed within the analysis window, where the matrix elements are obtained by weighted summation of the temporal cost, morphological cost, and response characteristic cost between ejection events and response events. The Hungarian algorithm is used to solve the cost matrix to obtain the optimal event matching sequence. Both ejection events and response events have a one-to-one correspondence with the cardiac cycle. The temporal cost is obtained based on the difference in R-wave peak timestamps between the ejection event and the response event; the morphological cost is obtained based on the difference in pulse pressure between the ejection event and the response event; and the response characteristic cost is obtained based on the difference in response time constants between the ejection event and the response event.

[0007] Furthermore, the method for obtaining the structural disorder index includes: The first structural index is determined based on the degree of inconsistency between the indexes of ejection events and response events in the optimal event matching sequence of cardiac cycles; the second structural index is determined based on the degree of positional deviation between the cardiac cycles in the optimal event matching sequence and the global pulsation index in the analysis window. By combining the first and second structural indicators, a structural disorder indicator is obtained.

[0008] Furthermore, the method for obtaining the functional disorder indicators includes: Based on the optimal event matching sequence, the pulse pressure sequence and response time constant sequence of the matched cardiac cycle are extracted; linear regression analysis is performed on the pulse pressure sequence and response time constant sequence to obtain the coefficient of determination; and functional disorder index is obtained through the coefficient of determination.

[0009] Furthermore, the aggregate time-series structural disorder indicators and functional disorder indicators include: During the preset observation period, the trend of structural disorder indicators and functional disorder indicators were obtained respectively; the mean and trend of structural disorder indicators during the preset observation period were weighted and summed to obtain a comprehensive score of structural disorder; the mean and trend of functional disorder indicators during the preset observation period were weighted and summed to obtain a comprehensive score of functional disorder. The binary pair consisting of the structural disorder comprehensive score and the functional disorder comprehensive score is used as the state assessment pair.

[0010] Furthermore, the method for obtaining the probability of hypotension events includes: The state assessment pair is used as input to a pre-trained risk assessment model, which outputs the predicted probability of hypotension events.

[0011] Furthermore, the early warning based on the probability of low blood pressure events includes: When the probability of a hypotension event exceeds the preset warning threshold, an early warning will be issued and the probability of a hypotension event and the status assessment will be displayed.

[0012] Furthermore, the method for obtaining the response time constant includes: The diastolic pressure waveform data for each cardiac cycle within the analysis window were extracted; the diastolic pressure waveform was decoupled by exponential decay fitting based on the binary Windkessel physiological model to obtain the corresponding response time constant.

[0013] Furthermore, the step of extracting a recent pulsation analysis window from the stroke-by-stroke hemodynamic recording sequence includes: With a preset time length and sliding step size, continuous recording segments are periodically extracted from the end of the beat-by-beat hemodynamic recording sequence; if the number of cardiac cycles contained in the extracted recording segment is greater than or equal to the preset minimum calculation threshold, the corresponding recording segment is determined as the analysis window.

[0014] The present invention has the following beneficial effects: This invention transforms coupling state assessment into a combinatorial optimization problem of finding the optimal event matching by constructing a ventricular-arterial event matching cost matrix and solving for the optimal matching sequence. It can calculate the pulse timing matching deviation and quantify the degree of structural disorder in the causal sequence of ventricular ejection and arterial response. This allows for earlier risk signals when the causal matching relationship is disrupted but the overall parameter fluctuation amplitude has not yet significantly increased. Based on corrected timing matching, structural disorder indicators are used to capture causal disorder before macroscopic parameters deteriorate, and functional dysfunction indicators are used to quantify the stability of system gain regulation, enabling specific differentiation between risk sources stemming from timing disorder and gain regulation failure. By quantifying structural disorder and functional dysfunction indicators in the ventricular-arterial coupling relationship, this invention analyzes the probability of hypotension events and provides early warnings, achieving precise quantitative early warning of early hemodynamic abnormalities. This effectively improves the comprehensiveness and efficiency of monitoring and provides reliable data support for precise clinical intervention. Attached Figure Description

[0015] Figure 1The flowchart illustrates a method for intelligent assessment and early warning of liver cirrhosis complications provided in one embodiment of the present invention. Detailed Implementation

[0016] Please see Figure 1 The diagram illustrates a flowchart of an intelligent assessment and early warning method for the risk of complications of cirrhosis provided by an embodiment of the present invention. The method includes the following steps: S1: Acquire electrocardiogram and arterial blood pressure signals synchronously in time sequence, identify the cardiac cycle and assign a global pulsation index, extract key features of the arterial blood pressure waveform, and generate a beat-by-beat hemodynamic recording sequence.

[0017] In this embodiment of the invention, continuous electrocardiogram (ECG) signals and arterial blood pressure signals can be acquired simultaneously. The sampling frequency can be set to 250Hz, which can be adjusted by the implementer according to the specific implementation scenario. Since the functional activities of the cardiovascular system are based on the cardiac cycle as the basic repetitive unit, continuous physiological signals can be divided into independent analysis units representing individual heartbeats, facilitating subsequent correlation analysis. A QRS complex detection algorithm can be used to determine the cardiac cycle, and a global beat index is introduced. The global beat index is a monotonically increasing integer starting from 1, used to uniquely identify the cardiac cycle, that is, to assign a unique global identity to each cardiac cycle. It should be noted that the method for determining the cardiac cycle is a well-known technique to those skilled in the art, and will not be limited or elaborated here.

[0018] Since the morphological changes in arterial blood pressure data during each cardiac cycle directly reflect key hemodynamic events such as ventricular ejection, arterial diastole, and aortic valve closure, further extraction of event features from the arterial blood pressure waveform segments of each cardiac cycle provides a data foundation for subsequent assessment of the ventricular-arterial interaction.

[0019] In this embodiment of the invention, for each cardiac cycle, the peak of the R wave is obtained by combining the QRS complex detection algorithm and peak detection on the electrocardiogram signal, and the time point corresponding to each R wave peak is recorded as the R wave peak timestamp. Further, the global maximum and global minimum values ​​are determined by traversing and searching the arterial blood pressure waveform. The pressure value corresponding to the global maximum value is taken as the systolic blood pressure value, the time point corresponding to the global maximum value is taken as the systolic blood pressure peak timestamp, and the pressure value corresponding to the global minimum value is taken as the diastolic blood pressure value.

[0020] In addition, the descending isthmus notch point in each cardiac cycle is located, which represents the aortic valve closure and the beginning of diastole. Specifically, within the time range after the peak systolic blood pressure timestamp, the arterial blood pressure waveform data is first subjected to Savitzky-Golay smoothing filtering, and then the numerical second derivative is calculated. The time point corresponding to the first local maximum point of the second derivative within this range is taken as the descending isthmus notch timestamp.

[0021] Finally, the R-wave peak timestamp, systolic blood pressure peak timestamp, and descending isthmus notch timestamp were used as the key timestamp set, and the difference between systolic and diastolic blood pressure values ​​was used as pulse pressure. The arterial blood pressure signal from the descending isthmus notch timestamp to the R-wave peak timestamp of the next cardiac cycle was used as diastolic waveform data to reflect the pressure decay process of the arterial system during diastole.

[0022] Each cardiac cycle encapsulates the global pulsation index, key timestamp set, pulse pressure, and diastolic waveform data to generate a single-stroke hemodynamic record, fully reflecting the hemodynamic characteristics of a single cardiac cycle. The single-stroke hemodynamic records are arranged according to the global pulsation index to obtain a stroke-by-stroke hemodynamic record sequence, reflecting the ventricular-arterial interaction process over continuous time.

[0023] S2: Extract the recent pulsation analysis window from the beat-by-beat hemodynamic recording sequence, decouple the diastolic waveform of each cardiac cycle in the analysis window from the physiological model, and extract the arterial system response time constant; construct the ventricular-arterial event matching cost matrix containing time cost, morphological cost and response characteristic cost, and solve to obtain the optimal event matching sequence.

[0024] By decoupling the diastolic waveform of arterial blood pressure during each cardiac cycle, the arterial system response time constant, reflecting the instantaneous load characteristics of the arterial system, is quantified. Based on this, a ventricular-arterial event matching cost matrix is ​​constructed and solved, thereby calculating the pulsatility timing matching deviation and ventricular ejection-arterial response gain instability, which respectively characterize the degree of disorder in the causal relationship in two dimensions: timing matching and gain stability. This enables early quantification of the decompensation process in the circulatory system.

[0025] Because a patient's hemodynamic state can change rapidly within seconds to minutes, continuous and dynamic analysis of recent data is necessary. In this embodiment of the invention, continuous recording segments are periodically extracted from the end of the beat-by-beat hemodynamic recording sequence with a preset time length and sliding step size. End-to-end extraction ensures that the analysis focuses on the latest state, while the sliding step size balances real-time performance and computational efficiency. Specifically, with a preset time length, such as 15 seconds, as the analysis window length and a preset sliding step size, such as 5 seconds, a segment containing several recent heartbeat recordings is extracted from the end of the beat-by-beat hemodynamic recording sequence as a candidate analysis window.

[0026] To ensure statistical reliability, if the number of cardiac cycles contained in the extracted recording segment is greater than or equal to the preset minimum calculation threshold, the corresponding recording segment is designated as the analysis window; otherwise, the calculation of the current segment is skipped, and the next cycle segment is analyzed to ensure data validity. In this embodiment of the invention, the preset minimum calculation threshold can be set to 5. It should be noted that the time length, sliding step size, and minimum threshold can be adjusted by the implementer according to clinical monitoring needs and the characteristics of patient condition fluctuations, and are not absolutely limited here.

[0027] One of the core processes of circulatory decompensation after TIPS is systemic vascular tone imbalance. This imbalance directly alters the arterial system load characteristics, which are jointly determined by total peripheral resistance and total arterial compliance. These characteristics are directly reflected in the rate of diastolic pressure decay of arterial blood pressure. Therefore, the diastolic waveform of each heartbeat within the analysis window is fitted to decouple and quantify the arterial system response time constant of this decay rate.

[0028] In this embodiment of the invention, diastolic pressure waveform data for each cardiac cycle within the analysis window are extracted. Considering the potential boundary effect of incomplete data at the end of the window, in one specific embodiment, cardiac cycles corresponding to the global pulsation index at the beginning and end of the analysis window are not analyzed, while the remaining cardiac cycles are analyzed as a whole. The diastolic pressure waveform is decoupled by exponential decay fitting based on a binary Windkessel physiological model to obtain the corresponding response time constant, reflecting the elastic recoil capacity and peripheral resistance state of the arterial system.

[0029] In this embodiment of the invention, the exponential decay model is linearized into a logarithmic form, and the time variable is converted into a relative time with respect to the start time of the diastolic pressure waveform data. The fitting slope is calculated through linear regression, and the response time constant is the negative value of the reciprocal of the fitting slope, i.e. ; Represented as the response time constant, This is represented as the fitted slope.

[0030] Considering potential artifact interference, if the calculated fitting slope is greater than or equal to -10... -5 If the pressure does not effectively decrease, the response time constant is set to a preset upper limit value. In this embodiment of the invention, the preset upper limit value can be set to 3 seconds, which can be adjusted by the implementer and is not limited here.

[0031] Because healthy ventricular-arterial coupling is characterized by a high degree of beat-by-beat consistency between ventricular ejection (characterized by pulse pressure) and arterial response (characterized by response time constant), and decompensation processes disrupt this consistency, a cost matrix that can comprehensively assess this multi-dimensional matching distortion is constructed to provide a quantitative basis for subsequent identification of intrinsic causal disorders.

[0032] In one specific embodiment of the present invention, a ventricular-arterial event matching cost matrix is ​​constructed within the analysis window, wherein the ejection event index and the response event index correspond to cardiac cycles arranged chronologically within the analysis window, and the number of rows and columns of the matrix are both the number of valid cardiac cycles within the window. Both ejection events and response events have a one-to-one correspondence with cardiac cycles.

[0033] In one specific implementation, assuming the analysis window contains n cardiac cycles, a ventricular-arterial event matching cost matrix C of size n×n is constructed. The matrix element C(i,j) represents the total cost required to match the i-th ejection event with the j-th response event.

[0034] First, the time series cost is obtained based on the difference in R-wave peak timestamps between the ejection event and the response event. Specifically, the absolute difference between the two R-wave peak timestamps is calculated, and the ratio of the absolute difference in timestamps to the average cardiac cycle length within the analysis window is used as the time series cost. The greater the difference, the higher the time series cost.

[0035] Furthermore, the morphological cost is obtained based on the difference in pulse pressure between the ejection event and the response event. Specifically, the absolute difference in pulse pressure between the two events is calculated, and the ratio of the absolute difference in pulse pressure to the average pulse pressure within the analysis window is used as the morphological cost. The greater the difference, the higher the morphological cost.

[0036] Furthermore, the response characteristic is obtained based on the difference in response time constants corresponding to the ejection event and the response event. Specifically, the absolute difference between the two response time constants is calculated, and the ratio between the absolute difference of the response time constant and the average response time constant within the analysis window is used as the response characteristic, which reflects the matching deviation of arterial load characteristics. The greater the difference, the higher the response characteristic.

[0037] The matrix elements are obtained by weighted summation of the temporal cost, morphological cost, and response characteristic cost between the ejection event and the response event. As an example, the expression for each element of the matrix is: In the formula, Represented as time series cost, Represented as morphological cost, Represented as the cost of response characteristics, , and The weights are represented as weighted values, which can be set to 0.4, 0.4 and 0.2 in this embodiment of the invention. The implementer can adjust them as needed, and there are no restrictions here.

[0038] Furthermore, the Hungarian algorithm is used to solve the cost matrix. The Hungarian algorithm can find the solution with the minimum total cost among all one-to-one matching relationships, obtaining the optimal event matching sequence. This sequence represents a one-to-one correspondence mapping the blood ejection event index to the response event index, minimizing the total cost of all matching pairs and reflecting the most reasonable potential causal pairing relationship within the system. It should be noted that the solution process of the Hungarian algorithm is a well-known prior art, and this embodiment only utilizes its optimal matching solution capability, which will not be elaborated upon here.

[0039] Thus, the extraction of the optimal matching structure reflecting the deep coupling state of the system from the original waveform has been completed.

[0040] S3: Determine the structural disorder index based on the difference between the optimal event matching sequence and the natural time series; determine the functional disorder index by analyzing the correlation stability between pulse pressure and response time constant under the optimal event matching sequence.

[0041] Based on the optimal event matching sequence, in-depth analysis was conducted to consider that the hemodynamic instability after TIPS may be caused by causal timing disorder or modulation gain failure. The degree of disorder in ventricular-arterial coupling was quantified from two dimensions. The pulsation timing deviation and response gain instability were analyzed by measuring the matching deviation between sequences and the stability of linear correlation.

[0042] In this embodiment of the invention, since the pairing of ventricular ejection and arterial response in a natural time series follows the order of the global pulsation index (i.e., ejection in the k-th cardiac cycle corresponds to response in the k-th cardiac cycle), and the optimal event matching sequence reflects the true causal pairing relationship, structural disorder can be quantified by comparing the differences between the two. Based on the degree of inconsistency between the indices of ejection and response events in the optimal event matching sequence, a first structural index is determined to characterize the frequency of time-series offset events. Specifically, the proportion of cardiac cycles in the optimal event matching sequence where the matched response event index is not equal to the ejection event index is statistically analyzed; this proportion represents the degree of difference between the matched sequence and the original natural sequence, serving as the first structural index. The higher the first structural index, the more causal misalignment pulsation events occur within the system.

[0043] The second structural index is determined based on the degree of deviation between the cardiac cycle position in the optimal event matching sequence and the global pulsation index in the analysis window. This index characterizes the severity of temporal mismatch. Specifically, the average absolute difference between the ejection event index and the response event index of all matching pairs in the optimal event matching sequence is calculated as the second structural index. The larger the second structural index, the greater the time span between ejection and response, and the more severe the causal delay or preemption phenomenon.

[0044] Finally, by combining the first and second structural indicators, a structural disorder indicator is obtained to comprehensively quantify structural disorder. In this embodiment of the invention, the first and second structural indicators are weighted and summed to obtain the structural disorder indicator, where the weights can be set to 0.5 and 0.5, and the specific values ​​can be adjusted by the implementer.

[0045] Furthermore, according to the principles of cardiovascular physiology, within a steady-state period, an increase in ventricular ejection energy, manifested as an increase in pulse pressure, should cause an approximately linear change in the arterial system's response characteristics, i.e., the response time constant. This linear relationship represents the stability of the coupled system gain.

[0046] Therefore, in this embodiment of the invention, based on the optimal event matching sequence, the pulse pressure sequence and response time constant sequence of the matched cardiac cycle are extracted to construct corrected causal data pairs, i.e., the data point corresponding to each data pair is (pulse pressure, response time constant). Linear regression analysis is performed on all data pairs to obtain the coefficient of determination. In a specific embodiment of the invention, the standard univariate linear regression algorithm is used to calculate the coefficient of determination, which is expressed as R0. 2 Its value ranges from [0, 1]. The closer it is to 1, the stronger the linear relationship and the more stable the gain.

[0047] Therefore, a functional impairment index is obtained through the coefficient of determination. In this embodiment of the invention, the difference between the value 1 and the coefficient of determination is used as the functional impairment index to reflect the degree of disruption of the linear gain relationship between ventricular ejection and arterial response. As an example, the expression is: , Indicators of functional dysfunction.

[0048] S4: Aggregate time-series structural disorder indicators and functional disorder indicators, obtain the risk probability of hypotension events through a pre-trained risk assessment model, and provide early warning based on the risk probability of hypotension events.

[0049] Since the decompensation of the circulatory system is a cumulative evolutionary process, the risk depends not only on the current state of the indicators, but also on the aggregated evaluation of the structural disorder indicators and functional disorder indicators over time to characterize the dynamic behavioral statistical features over a period of time.

[0050] In this embodiment of the invention, within a preset observation period, the trend of structural disorder indicators and functional disorder indicators are obtained respectively, reflecting the direction of deterioration or improvement of the indicators in the near future. Specifically, a linear fit is performed on continuous structural disorder indicators during the observation period, and the slope is extracted to obtain the trend of the structural disorder indicators. Similarly, a linear fit is performed on continuous functional disorder indicators during the observation period, and the slope is extracted to obtain the trend of the functional disorder indicators. A positive slope indicates that the indicator is showing a continuous deterioration trend, while a negative slope indicates that the indicator is showing an improvement trend. The preset observation period can be set to 15 minutes, and its length can be flexibly adjusted according to clinical early warning needs, without limitation. It should be noted that linear fitting is a technique well known to those skilled in the art, such as the least squares method, and will not be elaborated here.

[0051] Further weighted summation of the mean and trend of structural disorder indicators during the preset observation period yielded a comprehensive structural disorder score. A higher comprehensive structural disorder score indicates a more severe physiological misalignment between each ventricular ejection event and its temporally most closely matched arterial response event, with the misalignment showing a recent trend of intensification. As an example, the expression for the comprehensive structural disorder score is: In the formula, This is represented as a comprehensive score for structural disorder. This is represented as the mean of the structural disorder index. The trend degree is represented as an indicator of structural disorder. and The weights, represented as weighted values, can be set to, for example, 0.6 and 0.4 in this embodiment of the invention, so that the sum of the weights is 1. The actual values ​​can be obtained by training on historical clinical data, etc., and the implementer can adjust them at his own discretion, without any restrictions.

[0052] Further weighted summation of the mean and trend of functional dysfunction indicators during the preset observation period yielded a comprehensive functional dysfunction score. A higher comprehensive functional dysfunction score indicates a more unstable linear causal relationship between ventricular ejection intensity and arterial response characteristics, with the instability showing a recent trend of intensification, reflecting the degree of dysfunction in the circulatory system's gain regulation mechanism. As an example, the expression for the comprehensive functional dysfunction score is: In the formula, This is represented by a comprehensive functional disorder score. This is expressed as the mean of a functional disorder index. This is expressed as the trend of a functional disorder indicator.

[0053] A binary pair consisting of a structural disorder comprehensive score and a functional disorder comprehensive score is used as a state assessment pair. The predicted probability of a hypotensive event is jointly assessed using these state assessment pairs. In this embodiment of the invention, the pre-trained risk assessment model is a logistic regression model trained based on a large amount of historical patient data. During training, the state assessment pairs of historical patients are used as input, and the occurrence of a hypotensive event within a preset prediction time window, such as 30 minutes, is used as a label. The criterion for a hypotensive event is a mean arterial pressure below 65 mmHg for 5 consecutive minutes. The model parameters are optimized through algorithmic iteration, ultimately resulting in a pre-trained model with fixed parameters. This model has the advantages of simple structure and interpretable results. Therefore, the currently generated state assessment pair is used as input to the risk assessment model, outputting the predicted probability of a hypotensive event. This probability ranges from [0,1], with higher values ​​indicating a greater likelihood of a future hypotensive event.

[0054] This invention provides early warning based on the probability of hypotension events. In this embodiment, when the probability of a hypotension event exceeds a preset warning threshold, an alarm mechanism is triggered to issue a warning and display the probability of hypotension events and a status assessment. The preset warning threshold can be set to 0.7, which can be adjusted by the implementer. The status assessment intuitively reflects the current level of structural disorder and functional dysfunction, providing clinical medical staff with clear risk quantification results and guidance for intervention.

[0055] In summary, this invention transforms the assessment of coupling state into a combinatorial optimization problem of finding the optimal event matching by constructing a ventricular-arterial event matching cost matrix and solving for the optimal matching sequence. This allows for the calculation of pulsation timing matching deviations and quantifies the degree of structural disorder in the causal sequence of ventricular ejection and arterial response. This provides earlier risk signals even when the causal matching relationship is disrupted but the overall parameter fluctuation amplitude has not yet significantly increased. Based on corrected timing matching, structural disorder indicators capture causal disorder before macroscopic parameters deteriorate, and functional dysfunction indicators quantify the stability of system gain regulation, enabling specific differentiation between risk sources stemming from timing disorder and gain regulation failure. By quantifying structural disorder and functional dysfunction indicators in the ventricular-arterial coupling relationship, this invention analyzes the probability of hypotension events and provides early warnings, achieving precise quantitative early warning of early hemodynamic abnormalities. This effectively improves the comprehensiveness and efficiency of monitoring and provides reliable data support for precise clinical intervention.

Claims

1. A method for intelligent assessment and early warning of the risk of complications of cirrhosis, characterized in that, The method includes: Acquire electrocardiogram and arterial blood pressure signals synchronously in time sequence, identify cardiac cycles and assign global pulsation indexes, extract key features of arterial blood pressure waveforms, and generate a beat-by-beat hemodynamic recording sequence; The recent pulsation analysis window is extracted from the beat-by-beat hemodynamic recording sequence. The diastolic waveform of each cardiac cycle in the analysis window is decoupled from the physiological model, and the arterial system response time constant is extracted. The ventricular-arterial event matching cost matrix containing time cost, morphological cost and response characteristic cost is constructed, and the optimal event matching sequence is obtained by solving it. Based on the difference between the optimal event matching sequence and the natural time series, structural disorder indicators are determined; by analyzing the correlation stability between pulse pressure and response time constant under the optimal event matching sequence, functional disorder indicators are determined. By aggregating time-series structural disorder indicators and functional disorder indicators, the risk probability of hypotension events is obtained through a pre-trained risk assessment model; and early warning is provided based on the risk probability of hypotension events.

2. The intelligent assessment and early warning method for the risk of complications of cirrhosis according to claim 1, characterized in that, The beat-by-beat hemodynamic recording sequence is composed of single-beat hemodynamic recordings arranged in the order of the global pulsation index; each single-beat hemodynamic recording encapsulates data related to a single cardiac cycle, including a global pulsation index, a set of key timestamps, pulse pressure, and diastolic waveform data; the set of key timestamps includes the R-wave peak timestamp, the systolic blood pressure peak timestamp, and the descending isthmus notch timestamp, and the pulse pressure is the difference between the systolic blood pressure value and the diastolic blood pressure value.

3. The intelligent assessment and early warning method for the risk of complications of cirrhosis according to claim 2, characterized in that, The method for obtaining the optimal event matching sequence includes: A ventricular-arterial event matching cost matrix is ​​constructed within the analysis window, where the matrix elements are obtained by weighted summation of the temporal cost, morphological cost, and response characteristic cost between ejection events and response events. The Hungarian algorithm is used to solve the cost matrix to obtain the optimal event matching sequence. Both ejection events and response events have a one-to-one correspondence with the cardiac cycle. The temporal cost is obtained based on the difference in R-wave peak timestamps between the ejection event and the response event; the morphological cost is obtained based on the difference in pulse pressure between the ejection event and the response event; and the response characteristic cost is obtained based on the difference in response time constants between the ejection event and the response event.

4. The intelligent assessment and early warning method for the risk of complications of cirrhosis according to claim 3, characterized in that, The method for obtaining the structural disorder index includes: The first structural index is determined based on the degree of inconsistency between the indexes of ejection events and response events in the optimal event matching sequence of cardiac cycles; the second structural index is determined based on the degree of positional deviation between the cardiac cycles in the optimal event matching sequence and the global pulsation index in the analysis window. By combining the first and second structural indicators, a structural disorder indicator is obtained.

5. The intelligent assessment and early warning method for the risk of complications of cirrhosis according to claim 1, characterized in that, The methods for obtaining the functional disorder indicators include: Based on the optimal event matching sequence, the pulse pressure sequence and response time constant sequence of the matched cardiac cycle are extracted; linear regression analysis is performed on the pulse pressure sequence and response time constant sequence to obtain the coefficient of determination; and functional disorder index is obtained through the coefficient of determination.

6. The intelligent assessment and early warning method for the risk of complications of cirrhosis according to claim 1, characterized in that, The aggregated time-series structural disorder indicators and functional disorder indicators include: During the preset observation period, the trend of structural disorder indicators and functional disorder indicators were obtained respectively; the mean and trend of structural disorder indicators during the preset observation period were weighted and summed to obtain a comprehensive score of structural disorder; the mean and trend of functional disorder indicators during the preset observation period were weighted and summed to obtain a comprehensive score of functional disorder. The binary pair consisting of the structural disorder comprehensive score and the functional disorder comprehensive score is used as the state assessment pair.

7. The intelligent assessment and early warning method for the risk of complications of cirrhosis according to claim 6, characterized in that, The methods for obtaining the probability of hypotension events include: The state assessment pair is used as input to a pre-trained risk assessment model, which outputs the predicted probability of hypotension events.

8. The intelligent assessment and early warning method for the risk of complications of cirrhosis according to claim 7, characterized in that, The method of providing early warning based on the probability of low blood pressure events includes: When the probability of a hypotension event exceeds the preset warning threshold, an early warning will be issued and the probability of a hypotension event and the status assessment will be displayed.

9. The intelligent assessment and early warning method for the risk of complications of cirrhosis according to claim 1, characterized in that, The method for obtaining the response time constant includes: The diastolic pressure waveform data for each cardiac cycle within the analysis window were extracted; the diastolic pressure waveform was decoupled by exponential decay fitting based on the binary Windkessel physiological model to obtain the corresponding response time constant.

10. The intelligent assessment and early warning method for the risk of complications of cirrhosis according to claim 1, characterized in that, The step of extracting a recent pulsation analysis window from a beat-by-beat hemodynamic recording sequence includes: With a preset time length and sliding step size, continuous recording segments are periodically extracted from the end of the beat-by-beat hemodynamic recording sequence; if the number of cardiac cycles contained in the extracted recording segment is greater than or equal to the preset minimum calculation threshold, the corresponding recording segment is determined as the analysis window.

Images