A smart multi-parameter risk early warning system for critically ill patients

Abstract

This invention relates to the field of medical monitoring technology, specifically to an intelligent multi-parameter risk early warning system for critically ill patients. The system transforms ventricular amplitude and pulse pressure sequences within a preset sliding data window to determine relative amplitude and relative pulse pressure sequences; it then determines pulse pressure variability based on the pulse pressure difference sequence within the window; it determines a reference conduction time based on the electromechanical delay sequence within the window; it constructs a matching cost matrix based on the reference conduction time, relative amplitude, and relative pulse pressure sequences, and solves for the optimal matching relationship; it constructs an excitation-response data pair set based on the optimal matching relationship and determines a waveform difference index; it determines the response inversion ratio based on the set; and it performs graded screening based on pulse pressure variability, waveform difference index, and response inversion ratio to determine physiological status and output corresponding risk warnings, thereby significantly improving the ability to identify early cardiovascular abnormalities.

Description

Technical Field

[0001] This invention relates to the field of medical monitoring technology, specifically to an intelligent multi-parameter risk early warning system for critically ill patients. Background Technology

[0002] In the field of critical care, mean arterial pressure (MAP) is commonly used as the primary indicator for assessing circulatory stability, and vasoactive drugs are adjusted based on blood pressure readings to maintain hemodynamic stability. However, MAP only reflects the macroscopic homeostasis of the cardiovascular system and is insufficient to reveal the vascular's own regulatory functions and responsiveness under drug action.

[0003] In pathological conditions such as vascular stiffness, endothelial damage, or acidosis, even if blood pressure values ​​are still within the normal range, there may be a pathological state of poor coupling or even functional decoupling between cardiac electrical excitation and vascular mechanical response. At this time, the circulatory system is already on the verge of compensatory exhaustion, and existing monitoring methods are difficult to identify such early hidden abnormalities.

[0004] Moreover, existing monitoring methods mostly rely on blood pressure value trend analysis or single waveform comparison, which cannot eliminate the influence of individual differences and signal drift. They also lack quantitative assessment methods for the coupling relationship between cardiac electrical activity and vascular mechanical response, making it difficult to capture early warning signals before circulatory collapse, resulting in problems of delayed warning and inaccurate assessment. Summary of the Invention

[0005] To address the problem that existing critical care monitoring technologies, which rely solely on absolute blood pressure values, cannot identify early abnormalities in cardiovascular coupling, leading to delayed early warnings, this invention provides an intelligent multi-parameter risk early warning system for critically ill patients. The specific technical solution adopted is as follows: This invention proposes an intelligent multi-parameter risk early warning system for critically ill patients, the system comprising: The acquisition module is used to simultaneously acquire the target object's electrocardiogram (ECG) signal and invasive arterial pressure signal, and extract the ventricular amplitude, pulse pressure difference, and electromechanical delay for each cardiac cycle based on the R wave of the ECG signal. The transformation module is used to standardize the ventricular amplitude value sequence and pulse pressure difference sequence composed of all cardiac cycles within a preset sliding data window to determine the relative amplitude value sequence and relative pulse pressure value sequence; and to determine the pulse pressure variability based on the pulse pressure difference sequence within the window. The matrix module is used to determine the reference conduction time based on the electromechanical delay sequence within the window; construct a matching cost matrix based on the reference conduction time, the relative amplitude sequence, and the relative pulse compression sequence; where each element in the matrix represents the comprehensive cost required to pair one excitation with one response; and solve for the optimal matching relationship that minimizes the total matching cost based on the matrix. The determination module is used to construct an excitation-response data pair set based on the optimal matching relationship and determine the waveform difference index; and to determine the response inversion ratio based on the set. The early warning module is used to perform graded screening based on pulse pressure variability, waveform difference index and response reversal ratio, combined with preset thresholds, to determine the physiological state and output the corresponding risk warning.

[0006] Furthermore, the extraction of ventricular electrical amplitude, pulse pressure, and electromechanical delay for each cardiac cycle based on the R wave of the electrocardiogram signal includes: Identify the R-wave peaks of the electrocardiogram signal; define the time interval between two adjacent peaks in the continuously detected R-wave peaks as a cardiac cycle, and take the initial R-wave peak of each cardiac cycle as the reference. At the moment corresponding to the peak of the initial R wave, the voltage amplitude of the peak of the initial R wave relative to the isoelectric line of the cardiac cycle is extracted as the ventricular electrical amplitude of the cardiac cycle. Starting from the moment of the initial R wave peak, the corresponding arterial blood pressure waveform segment is identified from the invasive arterial pressure signal within a preset time search window. The peak systolic pressure and the trough diastolic pressure are extracted from the arterial blood pressure waveform segment, and the difference between the peak systolic pressure and the trough diastolic pressure is calculated as the pulse pressure difference of the cardiac cycle. Identify the onset point of systole in an arterial blood pressure waveform segment; calculate the time difference between the onset point of systole and the peak of the initial R wave as the electromechanical delay of the cardiac cycle.

[0007] Furthermore, the preset sliding data window operates in the following manner: At the initial stage of system startup, the ventricular electrical amplitude, pulse pressure difference, and electromechanical delay of each cardiac cycle are packaged into a set of cardiac cycle characteristic data and stored sequentially into the data window until the preset number of characteristic data sets are filled, thus completing the initialization of the data window; After the data window is full, each time a new set of cardiac cycle feature data is stored, the oldest set of feature data in the data window is removed to form an updated data window. For each updated data window, the system freezes a complete copy of the updated data window and performs subsequent standardization transformations, matrix construction, indicator and ratio calculations, and risk warnings based on the frozen copy.

[0008] Furthermore, the process of determining the relative amplitude value sequence and the relative pulse pressure value sequence includes: Calculate the arithmetic mean of the window ventricular electrical amplitude value sequence as the amplitude mean; calculate the standard deviation of the window ventricular electrical amplitude value sequence as the amplitude standard deviation. Calculate the arithmetic mean of the pulse pressure gradient sequence within the window as the pulse pressure mean; calculate the standard deviation of the pulse pressure gradient sequence within the window as the pulse pressure variability. For each cardiac cycle within a preset sliding data window, the sum of the standard deviation of the amplitude value and a preset minimum constant is calculated as the first sum; the first difference is divided by the first sum to obtain the relative amplitude value of the cardiac cycle; the second difference between the pulse pressure difference and the mean pulse pressure difference of the cardiac cycle is calculated; the sum of the pulse pressure variability and the preset minimum constant is calculated as the second sum; the second difference is divided by the second sum to obtain the relative pulse pressure value of the cardiac cycle. Traverse all cardiac cycles within the window to obtain a sequence of relative amplitude values ​​consisting of a preset number of relative amplitude values, and a sequence of relative pulse pressure values ​​consisting of a preset number of relative pulse pressure values.

[0009] Furthermore, the reference conduction time is the median of the electromechanical delay sequence within the window.

[0010] Furthermore, the matching cost matrix construction process includes: A two-dimensional array is constructed based on the relative amplitude value sequence and the relative pulse pressure value sequence. The rows in the array correspond to each excitation within the data window, and each value in the relative amplitude value sequence represents the intensity of the corresponding excitation. The columns in the array correspond to each response within the same data window, and each value in the relative pulse pressure value sequence represents the amplitude of the corresponding response. For each element in the two-dimensional array, calculate the absolute difference between the relative amplitude of the excitation at that element and the relative pulse pressure of the response at that element, and use it as the waveform difference value; The peak of the initial R-wave of the cardiac cycle corresponding to the element is taken as the actual time of the excitation at the element; the electromechanical delay of the corresponding cycle is added to the peak of the initial R-wave of the cardiac cycle corresponding to the element to obtain the actual time of the response at the element. The time difference between the actual occurrence time of the excitation at the element and the actual occurrence time of the response at the element is calculated as the actual conduction time; the absolute difference between the actual conduction time and the reference conduction time is calculated as the time difference; the sum of the reference conduction time and the preset minimum constant is calculated as the third sum; the time difference is divided by the third sum to obtain the time rationality coefficient; where, if the actual occurrence time of the response at the element is earlier than the actual occurrence time of the excitation at the element, the time rationality coefficient of the element is assigned to infinity; Based on preset weighting coefficients, the waveform difference value and the time rationality coefficient are weighted and summed to obtain the comprehensive cost. The comprehensive cost is then filled with the numerical values ​​of the elements; the completed two-dimensional array is the matching cost matrix.

[0011] Furthermore, the step of finding the optimal matching relationship that minimizes the total matching cost based on the matrix includes: The matching cost matrix is ​​solved using the Hungarian algorithm. Under the constraints that each excitation can only be paired with one response and each response can only be paired with one excitation, the global optimal solution that minimizes the sum of the comprehensive costs of all pairings is found. The optimal matching relationship is extracted based on the globally optimal solution. The optimal matching relationship is represented in the form of a mapping. For each excitation in the matrix, the column number of the response paired with the excitation is designated as the response number corresponding to the excitation, forming a one-to-one pairing relationship.

[0012] Furthermore, the process for determining the waveform difference index includes: Based on the optimal matching relationship, the relative amplitude value of each excitation and the relative pulse pressure value at the corresponding response number are combined into a set of excitation-response data pairs; the excitation-response data pairs are arranged in the time order of each excitation to form a set of excitation-response data pairs. Calculate the absolute difference between the relative amplitude and relative pulse pressure in each pair of excitation-response data as the degree of difference; Calculate the arithmetic mean of the differences in the excitation-response data pairs across all groups, and use this as the waveform difference index.

[0013] Furthermore, the response inversion ratio determination process includes: The data pairs are reordered in ascending order of their relative electric amplitude values ​​to generate a sorted set. Get the total number of all distinct position pairs in the sorted set; where, for a sorted set containing S data pairs, the total number of all distinct position pairs is S multiplied by S minus 1, and then divided by 2. After traversing the sorted set, find all pairs of positions where the previous position number is less than the next position number. For each pair, compare the relative pulse pressure value of the previous position with the relative pulse pressure value of the next position. If the relative pulse pressure value of the previous position is greater than the relative pulse pressure value of the next position, the pair is considered an inverted pair. Count the number of all reversed position pairs, and denote it as the reverse count; divide the reverse count by the total number of all distinct position pairs to obtain the response reverse ratio.

[0014] Furthermore, the step of performing graded screening based on pulse pressure variability, waveform difference index, and response inversion ratio, combined with a preset threshold, to determine the physiological state and output corresponding risk warnings includes: Upon entering the first level of screening, it is determined whether the pulse pressure variability is less than the preset fluctuation threshold; if so, the current state is determined to be a vascular unresponsive state and a high-risk warning is output, and no further screening is performed. If the first-level screening is not met, the second-level screening is initiated to determine whether the response inversion ratio is greater than the preset monotonicity threshold or whether the waveform difference index is greater than the preset waveform difference threshold. If either condition is met, the current state is determined to be a cardiovascular decoupling state and a medium-risk warning is output, and no further screening is performed. If the second-level screening does not meet the requirements, the third-level screening is initiated. If the pulse pressure variability is not less than the fluctuation limit threshold, the response inversion ratio does not exceed the monotonicity threshold, and the waveform difference index does not exceed the waveform difference threshold, the current state is determined to be a normal state of cardiovascular coupling, and a low-risk warning is issued.

[0015] The present invention has the following beneficial effects: This invention uses the ECG R wave as a reference to extract ventricular electrical amplitude, pulse pressure, and electromechanical delay within each cardiac cycle, achieving precise alignment between cardiac electrical excitation and vascular mechanical response. This avoids analytical errors caused by signal asynchrony and provides a stable data foundation for subsequent assessment of cardiovascular coupling status. By standardizing the ventricular electrical amplitude and pulse pressure sequences through a sliding data window, relative amplitude and relative pulse pressure sequences are obtained, reducing the impact of individual differences and baseline drift on the analysis results and ensuring the comparability of relative amplitude and relative pulse pressure values ​​obtained from different times. Based on reference conduction time and relative... A matching cost matrix is ​​constructed using electrical amplitude and relative pulse pressure values. Optimal matching is then used to achieve the globally optimal pairing relationship between excitation and response, enabling the objective identification of the reasonableness of the delay between electrical excitation and mechanical response. Based on this optimal matching relationship, the waveform difference index reflects the degree of waveform consistency between excitation and response, the response inversion ratio reflects the temporal orderliness of the electromechanical response, and pulse pressure variability reflects the fluctuation amplitude of the response. By comprehensively evaluating the quality of cardiovascular coupling from three dimensions—fluctuation amplitude, waveform consistency, and temporal orderliness—the ability to identify early abnormalities is significantly improved, enabling the capture of early warning signals. Attached Figure Description

[0016] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a schematic diagram of the structure of an intelligent multi-parameter risk early warning system for critically ill patients provided in one embodiment of the present invention; Figure 2 This is an example diagram illustrating the matching cost matrix construction process provided in one embodiment of the present invention. Detailed Implementation

[0018] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of an intelligent multi-parameter risk early warning system for critically ill patients proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0020] The following description, in conjunction with the accompanying drawings, details a specific scheme for an intelligent multi-parameter risk early warning system for critically ill patients provided by the present invention.

[0021] Please see Figure 1 The diagram illustrates a schematic of an intelligent multi-parameter risk early warning system for critically ill patients according to an embodiment of the present invention. The system includes: The acquisition module 101 is used to simultaneously acquire the electrocardiogram (ECG) signal and invasive arterial pressure signal of the target object, and extract the ventricular electrical amplitude, pulse pressure difference and electromechanical delay of each cardiac cycle based on the R wave of the ECG signal.

[0022] The electrocardiogram (ECG) and invasive arterial pressure (IPP) signals processed by this system are routinely acquired physiological parameters for clinical monitoring. The signal acquisition process follows standard medical operating procedures and is performed by qualified medical personnel in an appropriate medical environment. This system does not involve any form of medical intervention or operational guidance during the signal acquisition process; it only analyzes and processes the acquired signals. The target population refers to critically ill individuals receiving routine clinical monitoring. The output results of this system are only reference information to assist clinical decision-making and should not be used as the sole basis for direct diagnosis or treatment. The final clinical judgment should be made by professional medical personnel based on a comprehensive assessment of the patient's specific condition.

[0023] Synchronous acquisition refers to the simultaneous recording of electrocardiogram (ECG) signals and invasive arterial pressure signals of the same target object on the same time axis. In practice, the system simultaneously performs analog-to-digital conversion on the analog signals output from the ECG leads and arterial pressure sensors at the same sampling rate (e.g., 500Hz or 1000Hz) to ensure that the acquired ECG signals and invasive arterial pressure signals are strictly aligned in time.

[0024] Electrocardiogram (ECG) signals are records of potential changes on the body surface caused by the electrical activity of the heart. They are basic physiological parameters for routine clinical monitoring of heart rhythm and electrical activity.

[0025] Invasive arterial pressure signal is an intravascular pressure waveform measured in real time by a pressure sensor connected to a catheter inserted into a peripheral artery. It is a core parameter for monitoring hemodynamics in critical care.

[0026] It should be noted that the methods for acquiring electrocardiogram (ECG) signals and invasive arterial pressure signals are existing conventional techniques, which can be directly applied in this embodiment. For example, ECG signals are acquired by attaching standard limb or precordial lead electrodes to the body surface; invasive arterial pressure signals are continuously measured by connecting a pressure sensor and a monitor after arterial puncture and catheterization.

[0027] The R wave in an electrocardiogram (ECG) signal is the characteristic peak with the highest amplitude and steepest slope in the QRS complex.

[0028] It should be noted that the R-wave extraction method is a mature technology in existing ECG signal processing, and this embodiment can be directly applied. For example, the system uses classic QRS detection algorithms such as Pan-Tompkins to identify the occurrence position and peak time of each R-wave in the acquired ECG signal.

[0029] It's important to understand that every heartbeat is triggered by electrical activity. Under normal physiological conditions, the electrical impulse from the sinoatrial node is conducted to the ventricles, causing depolarization of the ventricular myocardial cells. This is represented by the QRS complex on an electrocardiogram (ECG), with the R wave being the most prominent marker of ventricular depolarization. Therefore, each detected R wave marks the occurrence of a ventricular electrical excitation event, that is, the electrical starting point of a heartbeat. Dividing the cardiac cycle based on the R wave essentially uses each electrical excitation event as a time anchor.

[0030] In this embodiment, the R-wave peak of the electrocardiogram signal is identified; the time interval between two adjacent peaks in the continuously detected R-wave peaks is defined as a cardiac cycle, and the initial R-wave peak of each cardiac cycle is used as a reference; at the moment corresponding to the initial R-wave peak, the voltage amplitude of the initial R-wave peak relative to the isoelectric line of the cardiac cycle is extracted as the ventricular electrical amplitude of the cardiac cycle; starting from the moment of the initial R-wave peak, the corresponding arterial blood pressure waveform segment is identified from the invasive arterial pressure signal within a preset time search window, the systolic blood pressure peak and diastolic blood pressure trough are extracted from the arterial blood pressure waveform segment, and the difference between the systolic blood pressure peak and diastolic blood pressure trough is calculated as the pulse pressure difference of the cardiac cycle; the starting point of the systolic phase in the arterial blood pressure waveform segment is identified; the time difference between the starting point of the systolic phase and the moment of the initial R-wave peak is calculated as the electromechanical delay of the cardiac cycle.

[0031] The isoelectric line refers to the baseline level of myocardial cells in an electrocardiogram (ECG) signal when they are in a resting state and without electrical activity. It is typically selected as a reference by the average amplitude of the PR segment (the smooth segment between the end of atrial depolarization and the beginning of ventricular depolarization) or the TP segment (the resting segment between the end of ventricular repolarization and the beginning of the next atrial depolarization). The method for determining the isoelectric line is a conventional technique in existing ECG signal processing and will not be elaborated upon in this embodiment.

[0032] It should be noted that identifying arterial blood pressure waveform segments and extracting the systolic blood pressure peak and diastolic blood pressure trough values ​​from these segments is common knowledge in the field and can be directly applied in this embodiment, so it will not be elaborated further.

[0033] Because there is a physiological delay in the mechanical contraction triggered by electrical signals, the vascular mechanical response induced by ventricular electrical excitation does not occur at the moment the R wave appears, but rather within a period of time after the R wave. The system uses the moment of the initial R wave peak as the starting point and sets a preset time search window to extract the pulse pressure difference.

[0034] It should be noted that the specific value of the preset time search window is determined based on medical experience, and this embodiment does not impose a specific limitation. For example, a typical value for the preset time search window is 50 to 400 milliseconds after the peak of the R wave.

[0035] It's important to understand that pulse pressure (PPP) refers to the difference between the peak systolic blood pressure and the trough diastolic blood pressure in an arterial blood pressure waveform. In clinical physiology, PPP directly reflects the combined effect of stroke volume and arterial compliance. According to the existing Frank-Starling heart law, increased ventricular contractility leads to increased stroke volume, which in turn increases systolic blood pressure and increases PPP; conversely, decreased contractility results in a smaller PPP. Simultaneously, PPP is also affected by vascular compliance. If the blood vessel walls are elastic, they can accommodate cardiac ejection and buffer pressure fluctuations; if the blood vessels are stiff or have excessive tension, the same stroke volume will produce a larger change in PPP. Therefore, PPP is an ideal parameter connecting cardiac electrical excitation (i.e., electrical excitation events characterized by the R wave) and vascular mechanical response. It reflects both the heart's pumping function and the vascular response characteristics, and has a clear physiological correspondence with ventricular electrical amplitude.

[0036] The starting point of the systolic phase in an arterial blood pressure waveform (also known as the foot point) refers to the inflection point where the pressure waveform transitions from a slow descent to a rapid rise during diastole. Foot point identification technology is a standard method in this field and falls within the scope of existing technology, so it will not be elaborated upon further. For example, a common method for identifying the foot point is to calculate the first derivative of the waveform and find the location where the derivative changes from positive to negative or from a small to a large abrupt change.

[0037] The transformation module 102 is used to perform standardized transformation on the ventricular amplitude value sequence and pulse pressure difference sequence composed of all cardiac cycles within a preset sliding data window to determine the relative amplitude value sequence and relative pulse pressure value sequence; and to determine the pulse pressure variability based on the pulse pressure difference sequence within the window.

[0038] As a preferred implementation, the preset sliding data window operates as follows: At the initial stage of system startup, the ventricular electrical amplitude, pulse pressure difference, and electromechanical delay of each cardiac cycle are packaged into a set of cardiac cycle feature data and stored sequentially into the data window until a preset number of feature data sets are stored, thus completing the initialization of the data window; after the data window is full, for each new set of cardiac cycle feature data stored, the earliest set of feature data in the data window is removed, forming an updated data window; for each updated data window, the system freezes a complete copy of the updated data window and performs subsequent standardization transformations, matrix construction, indicator and ratio calculations, and risk warnings based on the frozen copy.

[0039] A preset sliding data window is a first-in, first-out (FIFO) cache structure with a fixed length (i.e., a preset number) used to store feature data from multiple consecutive cardiac cycles. This data window always maintains the latest data from the preset number of sets: if a new set of cardiac cycle feature data is generated, it is stored at the end of the window, while the oldest set of data is automatically removed, ensuring that the data window always contains feature data from the most recent preset number of cardiac cycles.

[0040] The initial stage of system startup refers to the initial phase in which the system begins to run but has not yet completed data accumulation. In this stage, the system has just begun to collect electrocardiogram signals and invasive arterial pressure signals from the target object and extract characteristic data of the cardiac cycle. The number of data in the sliding data window has not yet reached the preset number.

[0041] It should be noted that the specific value of the preset quantity is determined by those skilled in the art through conventional experiments or experience, and this embodiment does not impose a specific limitation. For example, if the sample size is too small, the statistical quantities such as the mean and standard deviation will be too affected by random fluctuations, reducing the reliability of the analysis results; if the sample size is too large, the data window will contain too much historical information, resulting in a slow response. Therefore, a typical value for the preset quantity is 60, corresponding to a data window of 60 consecutive cardiac cycles.

[0042] It is important to understand that because ventricular electrical amplitude (usually measured in millivolts) and pulse pressure (usually measured in millimeters of mercury) belong to different physical dimensions, and because there are individual differences in baseline ECG amplitude and blood pressure levels among different individuals and even among the same individual under different conditions, directly comparing their absolute values ​​cannot reveal the intrinsic relationship between their fluctuation patterns. Therefore, it is necessary to transform the original physical quantities into a dimensionless relative sequence through standardization transformation, eliminating the influence of dimensions and individual differences, so that subsequent analysis can focus on the relative fluctuation patterns of electrical excitation and mechanical response in terms of energy form, rather than being constrained by the magnitude of absolute values.

[0043] In this embodiment, the arithmetic mean of the ventricular electrical amplitude value sequence within the window is calculated as the amplitude mean; the standard deviation of the ventricular electrical amplitude value sequence within the window is calculated as the amplitude standard deviation; the arithmetic mean of the pulse pressure difference sequence within the window is calculated as the pulse pressure mean; the standard deviation of the pulse pressure difference sequence within the window is calculated as the pulse pressure variability; for each cardiac cycle within a preset sliding data window, the sum of the amplitude standard deviation and a preset minimum constant is calculated as the first sum; the first difference is divided by the first sum to obtain the relative amplitude value of the cardiac cycle; the second difference between the pulse pressure difference of the cardiac cycle and the pulse pressure mean is calculated; the sum of the pulse pressure variability and the preset minimum constant is calculated as the second sum; the second difference is divided by the second sum to obtain the relative pulse pressure value of the cardiac cycle; all cardiac cycles within the window are traversed to obtain a relative amplitude value sequence composed of a preset number of relative amplitude values, and a relative pulse pressure value sequence composed of a preset number of relative pulse pressure values.

[0044] It should be noted that the specific value of the preset minimum constant is determined based on engineering experience, and this embodiment does not impose a specific limitation. For example, a typical value for the preset minimum constant is... .

[0045] Since the mechanical compliance of the blood vessel wall is directly reflected in the amplitude of pulse pressure fluctuations with heartbeat, if the blood vessel is elastic and has self-regulating ability, the pulse pressure will fluctuate accordingly with changes in the intensity of the heartbeat; however, if the blood vessel is in a rigid state caused by excessive drug control, the pulse pressure will tend to be constant with minimal fluctuations. Therefore, by quantitatively assessing the dispersion of the pulse pressure sequence as a physical indicator for evaluating vascular function, based on this physiological principle, the system calculates the standard deviation of the pulse pressure sequence within a window as the pulse pressure variability.

[0046] Pulse pressure variability reflects the degree of fluctuation in the amplitude of vascular mechanical response within a certain sliding data window.

[0047] Specifically, a larger pulse pressure variability within a sliding data window indicates a more significant difference in pulse pressure between different cardiac cycles and a wider range of fluctuations in vascular response amplitude. This means that the vascular wall maintains better elasticity and regulatory capacity, thus enabling it to respond and adjust more promptly to changes in cardiac impulse intensity. Conversely, a smaller pulse pressure variability within a sliding data window indicates that the pulse pressure is more constant and fluctuates less between different cardiac cycles. This means that the vascular wall is more likely to be in a rigid state and is increasingly losing its normal responsiveness to changes in cardiac impulse intensity.

[0048] Matrix module 103 is used to determine the reference conduction time based on the electromechanical delay sequence within the window; construct a matching cost matrix based on the reference conduction time, the relative amplitude value sequence, and the relative pulse pressure value sequence; where each element in the matrix represents the comprehensive cost required for pairing an excitation with a response; and solve for the optimal matching relationship that minimizes the total matching cost based on the matrix.

[0049] Because electromechanical delay in pathological conditions can deviate from the normal range due to abnormal events such as premature beats and conduction blocks, using the arithmetic mean as the reference conduction time would cause these outliers to significantly lower or raise the average level, resulting in the reference value deviating from the true conduction characteristics of most normal cycles. Therefore, the reference conduction time is the median of the electromechanical delay sequence within the window.

[0050] Reference conduction time refers to the typical level of electromechanical delay within a certain data window. It is used to characterize the representative conduction time required for an electrical signal to travel from cardiac excitation to the mechanical response of blood vessels under the current physiological state.

[0051] It is important to understand that, due to the drastic fluctuations in vascular tension under pathological conditions, the electromechanical conduction time may experience nonlinear drift. This means that a particular excitation in the natural time sequence (i.e., ventricular electrical excitation event, characterized by the R-wave peak in the electrocardiogram) may not correspond exactly to a response in the same sequence (i.e., the mechanical pulsation event generated by the blood vessel's electrical excitation of the heart, characterized by the pulse pressure difference in the arterial blood pressure waveform). If a direct comparison is performed one by one in chronological order, incorrect waveform matching results will occur due to time misalignment. Therefore, by constructing a matching cost matrix containing all possible excitation-response pairings, and assigning a comprehensive cost to each candidate pairing in the matrix, the problem of finding the correct correspondence is transformed into solving the global optimal matching problem that minimizes the total cost. This allows for the restoration of the most likely true correspondence between electrical excitation and mechanical response under the condition of allowing conduction time fluctuations.

[0052] Because there is a discrepancy between the number of excitation events and the number of response events within the data window, the matching cost matrix generated based on the two-dimensional array is not a square matrix, thus affecting the subsequent search for the globally optimal solution. Therefore, before constructing the matching cost matrix, virtual nodes can be introduced to expand the two-dimensional array into a square matrix, and the comprehensive cost of all pairings involving virtual nodes can be directly assigned the preset maximum penalty constant.

[0053] It should be noted that the specific value of the preset maximum penalty constant is determined based on engineering experience, and this embodiment does not impose a specific limitation. For example, a typical value of the preset maximum penalty constant can be set to 10 to 100 times the upper limit of the normal comprehensive cost. For instance, if the normal comprehensive cost does not exceed 20, the preset maximum penalty constant can be set to 200 or 2000.

[0054] The matching cost matrix construction process is as follows: Figure 2 As shown, it includes: S101-1: Construct a two-dimensional array based on the relative amplitude value sequence and the relative pulse pressure value sequence; wherein, the rows in the array correspond to each excitation within the data window, and each value in the relative amplitude value sequence represents the intensity of the corresponding excitation; the columns in the array correspond to each response within the same data window, and each value in the relative pulse pressure value sequence represents the amplitude of the corresponding response.

[0055] For example, assuming the preset quantity is represented by N, then if the relative amplitude value sequence contains N relative amplitude values, there are N excitations; if the relative pulse pressure value sequence contains N relative pulse pressure values, there are N responses. Therefore, the two-dimensional array is N rows and N columns, where each row corresponds to an excitation event within a data window, and each row uses the relative amplitude value of that excitation to characterize the intensity of that excitation; each column corresponds to a response event within the same data window, and each column uses the relative pulse pressure value of that response to characterize the amplitude of that response. The element in the i-th row and j-th column of the two-dimensional array represents the hypothetical case of pairing the i-th excitation with the j-th response.

[0056] S101-2: For each element in the two-dimensional array, calculate the absolute difference between the relative amplitude of the excitation at that element and the relative pulse pressure of the response at that element, and use it as the waveform difference value.

[0057] The waveform difference value refers to the absolute difference between the relative amplitude of an excitation and the relative pulse pressure of a response in a given excitation and its paired response. It is used to quantify the degree of matching between the morphology of the excitation and the response.

[0058] Among them, the smaller the waveform difference value, the closer the excitation intensity and response amplitude are in relative terms, that is, the more the electrical waveform and mechanical waveform match in shape; the larger the waveform difference value, the more significant the difference in relative terms between excitation intensity and response amplitude, that is, the more mismatched they are in shape.

[0059] S101-3: Take the peak of the initial R wave of the cardiac cycle corresponding to the element as the actual time of the excitation at the element; add the electromechanical delay of the corresponding cycle to the peak of the initial R wave of the cardiac cycle corresponding to the element to obtain the actual time of the response at the element.

[0060] For example, in the matching cost matrix, each element is uniquely determined by row index i and column index j, where row index i corresponds to the i-th cardiac cycle within the data window and column index j corresponds to the j-th cardiac cycle within the data window. Then, the excitation at element i refers to the excitation event of the i-th cardiac cycle, the intensity of which is characterized by the relative electrical amplitude of that cycle, and the occurrence time is the peak of the initial R-wave of that cycle. The response at element j refers to the response event of the j-th cardiac cycle, the amplitude of which is characterized by the relative pulse pressure of that cycle, and the occurrence time is the peak of the initial R-wave of that cycle plus the electromechanical delay of that cycle.

[0061] It should be noted that, using the above example, the cardiac cycle corresponding to the excitation at element i is the i-th cycle, and its actual occurrence time is the peak of the initial R-wave of the i-th cycle; the cardiac cycle corresponding to the response at element j is the j-th cycle, and its actual occurrence time is the peak of the initial R-wave of the j-th cycle plus the electromechanical delay of the j-th cycle. Therefore, the element in the i-th row and j-th column of the matrix evaluates the hypothetical case of pairing the i-th excitation with the j-th response, where the excitation and response come from different cardiac cycles, and i and j may be equal or unequal. This is the core significance of the matching cost matrix, enabling cross-cycle pairing of excitation and response to address the temporal drift problem caused by conduction time drift in pathological states.

[0062] S101-4: Calculate the time difference between the actual occurrence time of the excitation at the element and the actual occurrence time of the response at the element, as the actual conduction time; calculate the absolute difference between the actual conduction time and the reference conduction time, as the time difference; calculate the sum of the reference conduction time and the preset minimum constant, as the third sum; divide the time difference by the third sum to obtain the time rationality coefficient; wherein, if the actual occurrence time of the response at the element is earlier than the actual occurrence time of the excitation at the element, the time rationality coefficient of the element is assigned to infinity.

[0063] It should be noted that the conduction of electrical signals and the initiation of mechanical contraction both require time. The reference conduction time is the median of the electromechanical delay within the data window. The electromechanical delay itself is the physiological time required for an electrical signal to travel from the heart to the blood vessels and produce a mechanical effect. Therefore, even under extreme pathological conditions, the reference conduction time cannot be zero or close to zero; it must be a positive number.

[0064] It should be noted that those skilled in the art will understand that in actual computer implementation, "infinity" can be approximated by a preset maximum penalty constant, which is much larger than the range of normal comprehensive cost, for example, it can be set to 10 to 100 times the upper limit of normal comprehensive cost.

[0065] The temporal rationality coefficient refers to the relative degree to which the actual conduction time of a pair (assuming the i-th excitation and the j-th response) deviates from the reference conduction time, and is used to quantify the temporal rationality of the pair.

[0066] Among them, the smaller the time rationality coefficient of a pair, the closer the actual transmission time is to the reference transmission time, that is, the more the pair conforms to the typical transmission characteristics of the data window in terms of time; the larger the time rationality coefficient of a pair, the further the actual transmission time deviates from the reference transmission time, that is, the more unreasonable the pair is in terms of time.

[0067] S101-5: Based on the preset weight coefficients, the waveform difference value and the time rationality coefficient are weighted and summed to obtain the comprehensive cost. The comprehensive cost is then filled with the numerical values ​​of the elements; the completed two-dimensional array is the matching cost matrix.

[0068] It is understandable that if the time rationality coefficient is assigned an infinite value, regardless of the magnitude of the waveform difference, the combined cost of the corresponding pairing of excitation and response will approach infinity. This means that the pairing completely violates the physiological causality in time—the response occurs before the excitation, which is physiologically impossible. Therefore, this pairing will be automatically excluded from the subsequent solution for optimal matching relationships due to its excessive cost and will not become a candidate for the final matching scheme.

[0069] It should be noted that the specific values ​​of the preset weighting coefficients are determined based on engineering experience, and this embodiment does not impose specific limitations. For example, since the waveform difference value is the standardized difference and the time reasonableness coefficient is the relative deviation, their typical value ranges from 0 to several times the standard deviation, and they are of similar magnitude. Therefore, taking a weighting coefficient of 1.0 can make the waveform difference value and the time reasonableness coefficient have similar contribution weights in the overall cost, achieving a balance between waveform matching and time reasonableness.

[0070] Overall cost = waveform difference value + preset weighting coefficient × time rationality coefficient.

[0071] The overall cost incorporates information from two dimensions: the waveform difference value reflects the degree of morphological matching, and the temporal rationality coefficient reflects the degree of temporal rationality. Specifically, the smaller the overall cost of a pair, the more rational the pair is in both morphological and temporal dimensions, meaning the excitation and response in that pair are more likely to be a true correspondence. Conversely, the larger the overall cost of a pair, the more significant the mismatch in morphology or time, meaning the excitation and response in that pair are less likely to be a true correspondence.

[0072] To accurately find the optimal matching relationship that minimizes the total matching cost, as an example, the Hungarian algorithm is used to solve the matching cost matrix. Under the constraints that each excitation can only be paired with one response and each response can only be paired with one excitation, the globally optimal solution that minimizes the sum of the comprehensive costs of all pairings is sought. The optimal matching relationship is extracted based on the globally optimal solution. The optimal matching relationship is represented in the form of a mapping. For each excitation in the matrix, the column number of the response paired with the excitation is designated as the response number corresponding to the excitation, forming a one-to-one pairing relationship.

[0073] It should be noted that the Hungarian algorithm is a classic algorithm for solving the assignment problem, capable of finding a one-to-one matching scheme that minimizes the total cost in polynomial time. This system applies the Hungarian algorithm to solve the matching cost matrix, quickly finding the globally optimal matching scheme that minimizes the sum of the comprehensive costs of all pairings under the constraints that each excitation can only be paired with a response once and each response can only be paired with an excitation once. Since the Hungarian algorithm itself is existing technology, those skilled in the art can implement its specific calculation process based on common knowledge; its detailed mathematical principles will not be elaborated here.

[0074] The globally optimal solution is the one that minimizes the sum of the overall costs of all possible excitation-response pairings among all possible one-to-one excitation-response pairings.

[0075] The specific process of extracting the optimal matching relationship based on the global optimal solution is as follows: the system traverses each excitation, checks which column the response it is paired with in the global optimal solution, and uses the column number as the response number corresponding to that excitation.

[0076] For example, assuming the data window length (i.e., the preset number) is 3, the globally optimal solution obtained by the Hungarian algorithm is: excitation 1 is paired with response 3; excitation 2 is paired with response 1; and excitation 3 is paired with response 2. The extracted optimal matching relationship is: the response number corresponding to excitation 1 is 3; the response number corresponding to excitation 2 is 1; and the response number corresponding to excitation 3 is 2. Therefore, the optimal matching relationship can be represented in mapping form as follows: =3, =1, =2, where This represents the response number (i.e., column index) corresponding to the i-th excitation.

[0077] The determination module 104 is used to construct an excitation-response data pair set based on the optimal matching relationship and determine the waveform difference index; and to determine the response inversion ratio based on the set.

[0078] The process of determining waveform difference index includes the following steps: 1) Based on the optimal matching relationship, the relative amplitude value of each excitation and the relative pulse pressure value at the corresponding response number are combined into a set of excitation-response data pairs; the excitation-response data pairs are arranged in the time order of each excitation to form a set of excitation-response data pairs.

[0079] In this context, "according to the time sequence of each excitation" refers to following the original time sequence of the cardiac cycle.

[0080] Excitation-response data pairs correspond to a single heartbeat event, reflecting the relationship between the intensity of electrical excitation of that heartbeat and the amplitude of its corresponding mechanical response under optimal matching conditions.

[0081] 2) Calculate the absolute difference between the relative amplitude and the relative pulse pressure in each pair of excitation-response data as the degree of difference.

[0082] The degree of difference reflects the difference between the electrical excitation intensity and the corresponding mechanical response amplitude of a specific heartbeat event (i.e., a specific cardiac cycle) under optimal matching conditions. Specifically, the smaller the degree of difference for a specific heartbeat event, the better the electrical waveform matches the mechanical waveform; conversely, the larger the degree of difference for a specific heartbeat event, the more significant the difference between the electrical waveform and the mechanical waveform.

[0083] 3) Calculate the arithmetic mean of the differences between all groups of excitation-response data pairs, and use it as the waveform difference index.

[0084] Among them, the smaller the waveform difference index, the better the electrical waveform matches the mechanical waveform as a whole, and the better the cardiovascular coupling state; the larger the waveform difference index, the more significant the difference between the electrical waveform and the mechanical waveform, and the more likely there is a pathological state of morphological mismatch.

[0085] It is important to understand that under normal physiological conditions, the heart and blood vessels follow a monotonic regulatory law of "increased input leading to increased output," meaning that an increase in electrical excitation should be accompanied by a corresponding increase in the amplitude of the mechanical response. However, under pathological conditions, this regulatory law may be disrupted, resulting in a functional decoupling phenomenon where increased excitation leads to a weakened response. Therefore, the degree of disruption of the monotonicity between input and output is used to assess the state of cardiovascular logical coupling, and this quantification result is used as the response inversion ratio output.

[0086] Input enhancement refers to the increase in the intensity of ventricular electrical excitation, that is, the increase in electrical input energy with each heartbeat, which is quantified by relative electrical amplitude in this scheme; output enhancement refers to the increase in the amplitude of vascular mechanical response, that is, the increase in the pulsation intensity generated by the blood vessels in response to the electrical excitation of the heart, which is quantified by relative pulse pressure in this scheme.

[0087] The process of determining the response inversion ratio includes the following steps: 1) Sort the data pairs in ascending order of their relative electric amplitude values ​​to generate a sorted set.

[0088] In the sorted set, the first data pair corresponds to the heartbeat event with the smallest excitation intensity (i.e., the smallest relative amplitude), and the last data pair corresponds to the heartbeat event with the largest excitation intensity (i.e., the largest relative amplitude). Each data pair in the sorted set still contains the excitation intensity of the heartbeat event and its corresponding response intensity, but the order of the data pairs has changed from chronological order to order of increasing excitation intensity.

[0089] 2) Obtain the total number of all distinct position pairs in the sorted set; where, for a sorted set containing S data pairs, the total number of all distinct position pairs is S multiplied by S minus 1, and then divided by 2.

[0090] For example, suppose there are S pairs of data in the sorted set. Since any two different positions (i.e., two different data pairs) can be chosen to form a position pair, the total number of all possible position pairs is equal to the number of combinations of choosing any two different positions from the S pairs of data. .

[0091] 3) Traverse all positions in the sorted set that satisfy the condition that the previous position number is less than the next position number. For each position pair, compare the relative pulse pressure value corresponding to the previous position with the relative pulse pressure value corresponding to the next position. If the relative pulse pressure value of the previous position is greater than the relative pulse pressure value of the next position, the position pair is determined to be an inverted position pair.

[0092] 4) Count the number of all reversed position pairs, and denote it as the reverse count; divide the reverse count by the total number of all different position pairs to obtain the response reverse ratio.

[0093] For example, suppose the sorted set contains four data pairs, each containing excitation intensity (relative amplitude value) and response intensity (relative pulse pressure value). Since the excitation intensities are already sorted in ascending order, for ease of explanation, we will only use the response intensity (relative pulse pressure value) of each data pair as an example. Assume the sorted set (arranged in ascending order of excitation intensity) is (position number 1, relative pulse pressure value 0.8), (position number 2, relative pulse pressure value 0.3), and (position number 3, relative pulse pressure value 0.9). Next, we check all position pairs that satisfy the condition that the preceding position number is less than the following position number, i.e., (position number 1, position number 2), (position number 1, position number 3), (position number 2, position number 3), a total of 3 position pairs; for (position number... 1. For position number 2) position pair, if the pulse pressure value of the previous position (0.8) is greater than the pulse pressure value of the subsequent position (0.3), the position pair (position number 1, position number 2) is determined to be an inverted position pair; for position number 1, position number 3) position pair, if the pulse pressure value of the previous position (0.8) is less than the pulse pressure value of the subsequent position (0.9), the position pair (position number 1, position number 3) is not an inverted position pair; for position number 2, position number 3) position pair, if the pulse pressure value of the previous position (0.3) is less than the pulse pressure value of the subsequent position (0.9), the position pair (position number 2, position number 3) is not an inverted position pair, then the inverted count is 1.

[0094] It should be noted that the data window length is a preset number, meaning that the data window must contain data from at least one cardiac cycle. However, in this system, the preset number is much greater than 1. Therefore, the total number of all different position pairs must be a positive number.

[0095] The response inversion ratio reflects the proportion of abnormal phenomena in the sorted set after ranking by excitation intensity that violate the monotonically increasing law of "the stronger the excitation, the stronger the response." Specifically, a response inversion ratio closer to 0 indicates that the response intensity also generally shows a monotonically increasing trend from small to large in the order of excitation intensity, with very few inversion phenomena. This strongly suggests that the cardiovascular system strictly follows the regulatory law described by Frank-Starling's law, "increased input leads to increased output," indicating a better cardiovascular-cardiovascular logical coupling. Conversely, a response inversion ratio closer to 1 indicates a more prevalent inversion phenomenon, with the response intensity almost completely opposite to the excitation intensity distribution—that is, the stronger the excitation, the weaker the response. This strongly suggests that the cardiovascular system has lost its normal regulatory law, or even exhibited a completely reversed abnormal state, indicating a more severely disrupted cardiovascular-cardiovascular logical coupling.

[0096] The early warning module 105 is used to perform graded screening based on pulse pressure variability, waveform difference index and response reversal ratio, combined with preset thresholds, to determine the physiological state and output the corresponding risk warning.

[0097] In this embodiment, the system enters the first-level screening stage to determine whether the pulse pressure variability is less than a preset fluctuation threshold. If so, the current state is determined to be a vascular unresponsive state, and a high-risk warning is output, without further screening. If the first-level screening is not met, the system enters the second-level screening stage to determine whether the response inversion ratio is greater than a preset monotonicity threshold or whether the waveform difference index is greater than a preset waveform difference threshold. If either condition is met, the current state is determined to be a cardiovascular decoupling state, and a medium-risk warning is output, without further screening. If neither of the second-level screening conditions is met, the system enters the third-level screening stage. If the pulse pressure variability is not less than the fluctuation threshold, the response inversion ratio does not exceed the monotonicity threshold, and the waveform difference index does not exceed the waveform difference threshold, the current state is determined to be a normal cardiovascular coupling state, and a low-risk warning is output.

[0098] It should be noted that the specific value of the preset fluctuation threshold is determined based on the background noise level of commonly used invasive blood pressure sensors in clinical practice, and this embodiment does not impose a specific limitation. For example, a typical value for the preset fluctuation threshold is 1-2 mmHg.

[0099] It's important to understand that if the pulse pressure variability is less than the fluctuation threshold, it indicates that the fluctuation range of the pulse pressure difference between different cardiac cycles is below the sensor's background noise level. This means the blood vessel wall is in a state of extreme rigidity or is "frozen" by the medication, losing its normal elastic regulatory capacity. In this case, the system directly outputs a high-risk warning and terminates the subsequent cost matrix calculation for the current data window, determining the current state as a vascular unresponsive state, indicating to clinical medical staff that the blood vessel has lost its self-regulating ability.

[0100] It should be noted that the specific value of the preset waveform difference threshold is determined based on the statistical distribution of waveform difference index in healthy individuals or those in a stable disease phase, and this embodiment does not impose a specific limitation. For example, the 95th percentile of the statistical distribution of waveform difference index can be used as the preset waveform difference threshold, such as a preset waveform difference threshold value of 1.0.

[0101] It should be noted that the specific value of the preset monotonicity threshold is usually determined between completely ordered (i.e., the response inversion ratio is 0) and completely random (i.e., the response inversion ratio is 0.5), and this embodiment does not impose a specific limitation. For example, a typical value of the preset monotonicity threshold is 0.4.

[0102] It's important to understand that if the response reversal ratio exceeds the preset monotonicity threshold, it indicates that the input-output regulation of the cardiovascular system has been disrupted, resulting in a functional decoupling phenomenon where excitation is enhanced while the response is weakened. Similarly, if the waveform difference index exceeds the preset waveform difference threshold, it means that even considering the influence of conduction time drift, there are still significant morphological differences between the electrical and mechanical waveforms. If either of these two conditions is met, it indicates a significant logical or morphological abnormality in the cardiovascular system. In this case, the system determines the current state as cardiovascular decoupling and outputs a medium-risk warning, indicating to clinicians that although blood pressure readings may be acceptable, the coordination between the heart and blood vessels has become abnormal, and circulatory compensatory capacity is nearing depletion.

[0103] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0104] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. An intelligent multi-parameter risk early warning system for critically ill patients, characterized in that, The system includes: The acquisition module is used to simultaneously acquire the target object's electrocardiogram (ECG) signal and invasive arterial pressure signal, and extract the ventricular amplitude, pulse pressure difference, and electromechanical delay for each cardiac cycle based on the R wave of the ECG signal. The transformation module is used to standardize the ventricular amplitude value sequence and pulse pressure difference sequence composed of all cardiac cycles within a preset sliding data window to determine the relative amplitude value sequence and relative pulse pressure value sequence; and to determine the pulse pressure variability based on the pulse pressure difference sequence within the window. The matrix module is used to determine the reference conduction time based on the electromechanical delay sequence within the window; construct a matching cost matrix based on the reference conduction time, the relative amplitude sequence, and the relative pulse compression sequence; where each element in the matrix represents the comprehensive cost required to pair one excitation with one response; and solve for the optimal matching relationship that minimizes the total matching cost based on the matrix. The determination module is used to construct an excitation-response data pair set based on the optimal matching relationship and determine the waveform difference index; and to determine the response inversion ratio based on the set. The early warning module is used to perform graded screening based on pulse pressure variability, waveform difference index and response reversal ratio, combined with preset thresholds, to determine the physiological state and output the corresponding risk warning.

2. The intelligent multi-parameter risk early warning system for critically ill patients according to claim 1, characterized in that, The extraction of ventricular electrical amplitude, pulse pressure, and electromechanical delay for each cardiac cycle based on the R wave of the electrocardiogram signal includes: Identify the R-wave peaks of the electrocardiogram signal; define the time interval between two adjacent peaks in the continuously detected R-wave peaks as a cardiac cycle, and take the initial R-wave peak of each cardiac cycle as the reference. At the moment corresponding to the peak of the initial R wave, the voltage amplitude of the peak of the initial R wave relative to the isoelectric line of the cardiac cycle is extracted as the ventricular electrical amplitude of the cardiac cycle. Starting from the moment of the initial R wave peak, the corresponding arterial blood pressure waveform segment is identified from the invasive arterial pressure signal within a preset time search window. The peak systolic pressure and the trough diastolic pressure are extracted from the arterial blood pressure waveform segment, and the difference between the peak systolic pressure and the trough diastolic pressure is calculated as the pulse pressure difference of the cardiac cycle. Identify the onset point of systole in an arterial blood pressure waveform segment; calculate the time difference between the onset point of systole and the peak of the initial R wave as the electromechanical delay of the cardiac cycle.

3. The intelligent multi-parameter risk early warning system for critically ill patients according to claim 1, characterized in that, The preset sliding data window operates as follows: At the initial stage of system startup, the ventricular electrical amplitude, pulse pressure difference, and electromechanical delay of each cardiac cycle are packaged into a set of cardiac cycle characteristic data and stored sequentially into the data window until the preset number of characteristic data sets are filled, thus completing the initialization of the data window; After the data window is full, each time a new set of cardiac cycle feature data is stored, the oldest set of feature data in the data window is removed to form an updated data window. For each updated data window, the system freezes a complete copy of the updated data window and performs subsequent standardization transformations, matrix construction, indicator and ratio calculations, and risk warnings based on the frozen copy.

4. The intelligent multi-parameter risk early warning system for critically ill patients according to claim 3, characterized in that, The process of determining the relative amplitude value sequence and the relative pulse compression value sequence includes: Calculate the arithmetic mean of the window ventricular electrical amplitude value sequence as the amplitude mean; calculate the standard deviation of the window ventricular electrical amplitude value sequence as the amplitude standard deviation. Calculate the arithmetic mean of the pulse pressure gradient sequence within the window as the pulse pressure mean; calculate the standard deviation of the pulse pressure gradient sequence within the window as the pulse pressure variability. For each cardiac cycle within a preset sliding data window, calculate the first difference between the ventricular electrical amplitude value and the mean amplitude value of the cardiac cycle; calculate the sum of the standard deviation of the amplitude value and a preset minimum constant as the first sum value; divide the first difference by the first sum value to obtain the relative amplitude value of the cardiac cycle; calculate the second difference between the pulse pressure difference and the mean pulse pressure difference of the cardiac cycle; calculate the sum of the pulse pressure variability and a preset minimum constant as the second sum value; divide the second difference by the second sum value to obtain the relative pulse pressure value of the cardiac cycle. Traverse all cardiac cycles within the window to obtain a sequence of relative amplitude values ​​consisting of a preset number of relative amplitude values, and a sequence of relative pulse pressure values ​​consisting of a preset number of relative pulse pressure values.

5. The intelligent multi-parameter risk early warning system for critically ill patients according to claim 2, characterized in that, The reference conduction time is the median of the electromechanical delay sequence within the window.

6. The intelligent multi-parameter risk early warning system for critically ill patients according to claim 5, characterized in that, The matching cost matrix construction process includes: A two-dimensional array is constructed based on the relative amplitude value sequence and the relative pulse pressure value sequence. The rows in the array correspond to each excitation within the data window, and each value in the relative amplitude value sequence represents the intensity of the corresponding excitation. The columns in the array correspond to each response within the same data window, and each value in the relative pulse pressure value sequence represents the amplitude of the corresponding response. For each element in the two-dimensional array, calculate the absolute difference between the relative amplitude of the excitation at that element and the relative pulse pressure of the response at that element, and use it as the waveform difference value; The peak of the initial R-wave of the cardiac cycle corresponding to the element is taken as the actual time of the excitation at the element; the electromechanical delay of the corresponding cycle is added to the peak of the initial R-wave of the cardiac cycle corresponding to the element to obtain the actual time of the response at the element. The time difference between the actual occurrence time of the excitation at the element and the actual occurrence time of the response at the element is calculated as the actual conduction time; the absolute difference between the actual conduction time and the reference conduction time is calculated as the time difference; the sum of the reference conduction time and the preset minimum constant is calculated as the third sum; the time difference is divided by the third sum to obtain the time rationality coefficient; where, if the actual occurrence time of the response at the element is earlier than the actual occurrence time of the excitation at the element, the time rationality coefficient of the element is assigned to infinity; Based on preset weighting coefficients, the waveform difference value and the time rationality coefficient are weighted and summed to obtain the comprehensive cost. The comprehensive cost is then filled with the numerical values ​​of the elements; the completed two-dimensional array is the matching cost matrix.

7. The intelligent multi-parameter risk early warning system for critically ill patients according to claim 6, characterized in that, The method of finding the optimal matching relationship that minimizes the total matching cost based on the matrix includes: The matching cost matrix is ​​solved using the Hungarian algorithm. Under the constraints that each excitation can only be paired with one response and each response can only be paired with one excitation, the global optimal solution that minimizes the sum of the comprehensive costs of all pairings is found. The optimal matching relationship is extracted based on the globally optimal solution. The optimal matching relationship is represented in the form of a mapping. For each excitation in the matrix, the column number of the response paired with the excitation is designated as the response number corresponding to the excitation, forming a one-to-one pairing relationship.

8. The intelligent multi-parameter risk early warning system for critically ill patients according to claim 7, characterized in that, The process for determining the waveform difference index includes: Based on the optimal matching relationship, the relative amplitude value of each excitation and the relative pulse pressure value at the corresponding response number are combined into a set of excitation-response data pairs; the excitation-response data pairs are arranged in the time order of each excitation to form a set of excitation-response data pairs. Calculate the absolute difference between the relative amplitude and relative pulse pressure in each pair of excitation-response data as the degree of difference; Calculate the arithmetic mean of the differences in the excitation-response data pairs across all groups, and use this as the waveform difference index.

9. The intelligent multi-parameter risk early warning system for critically ill patients according to claim 8, characterized in that, The process for determining the response inversion ratio includes: The data pairs are reordered in ascending order of their relative electric amplitude values ​​to generate a sorted set. Get the total number of all distinct position pairs in the sorted set; where, for a sorted set containing S data pairs, the total number of all distinct position pairs is S multiplied by S minus 1, and then divided by 2. After traversing the sorted set, find all pairs of positions where the previous position number is less than the next position number. For each pair, compare the relative pulse pressure value of the previous position with the relative pulse pressure value of the next position. If the relative pulse pressure value of the previous position is greater than the relative pulse pressure value of the next position, the pair is considered an inverted pair. Count the number of all reversed position pairs, and denote it as the reverse count; divide the reverse count by the total number of all distinct position pairs to obtain the response reverse ratio.

10. The intelligent multi-parameter risk early warning system for critically ill patients according to claim 1, characterized in that, The process involves hierarchical screening based on pulse pressure variability, waveform difference index, and response reversal ratio, combined with preset thresholds, to determine physiological states and output corresponding risk warnings, including: Upon entering the first level of screening, it is determined whether the pulse pressure variability is less than the preset fluctuation threshold; if so, the current state is determined to be a vascular unresponsive state and a high-risk warning is output, and no further screening is performed. If the first-level screening is not met, the second-level screening is initiated to determine whether the response inversion ratio is greater than the preset monotonicity threshold or whether the waveform difference index is greater than the preset waveform difference threshold. If either condition is met, the current state is determined to be a cardiovascular decoupling state and a medium-risk warning is output, and no further screening is performed. If the second-level screening does not meet the requirements, the third-level screening is initiated. If the pulse pressure variability is not less than the fluctuation limit threshold, the response inversion ratio does not exceed the monotonicity threshold, and the waveform difference index does not exceed the waveform difference threshold, the current state is determined to be a normal state of cardiovascular coupling, and a low-risk warning is issued.

Images