A method and system for early warning of indwelling catheter blockage based on pressure waveform analysis

Abstract

The present application relates to a kind of based on pressure waveform analysis's indwelling catheter blockage early warning method and system, it is related to monitoring early warning technical field, including acquisition contains original pressure waveform data, instantaneous flow rate data and electrocardiogram synchronous signal data multisource monitoring data set;Original pressure waveform data is executed based on Daubechies wavelet base function four-layer multiscale decomposition processing, generates multiscale waveform component set;Combining patient basic information constructs catheter downstream vascular bed fluid mechanics impedance model, deduces and generates catheter downstream vascular bed impedance change index;Through sliding time window, dynamic baseline drift modeling is carried out, and adaptive catheter blockage risk benchmark threshold is set;Construct patient-specific risk factor graph structure, through recurrent neural network, time series prediction analysis is carried out to feature sequence, generates future risk evolution trend index;Generate three-level early warning signal.The present application has beneficial effect for significantly improving the sensitivity and specificity of catheter blockage early warning.

Description

Technical Field

[0001] This invention relates to the field of monitoring and early warning technology, and more specifically, to a method and system for early warning of indwelling catheter blockage based on pressure waveform analysis. Background Technology

[0002] Currently, monitoring of indwelling catheter occlusion mainly relies on regular clinical rounds and static pressure readings. Healthcare professionals assess catheter function by observing changes in pressure waveform damping or alarm triggering. This manual interpretation method has a significant lag, only identifying occlusion when severe catheter blockage occurs and hemodynamics changes are significant, thus missing the intervention window for early thrombus formation. While some automated monitoring systems can continuously record pressure values, they often employ fixed threshold alarm strategies, failing to consider individual patient differences and dynamic baseline drift due to physiological rhythms, resulting in a high false alarm rate.

[0003] While existing technologies attempt to extract features using time-domain statistical indicators or simple spectral analysis, they lack a refined deconstruction of the multi-scale physiological rhythms of pressure waveforms, fail to fully integrate multimodal data such as flow velocity and electrocardiogram, and cannot construct a dynamic impedance model for the catheter-vessel coupling system. More importantly, traditional methods are all static assessments and cannot predict the future evolution of occlusion risk. Clinical practice still relies on empirical periodic flushing maintenance, lacking data-driven, precise early warning mechanisms. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for early warning of indwelling catheter blockage based on pressure waveform analysis, so as to improve the above-mentioned problems. To achieve the above objective, the technical solution adopted by this invention is as follows:

[0005] In a first aspect, this application provides a method for early warning of indwelling catheter blockage based on pressure waveform analysis, including:

[0006] The hemodynamic multimodal sensing system, which integrates a distal pressure sensor of the indwelling catheter, a flow sensor of the catheter line, and a patient's electrocardiogram monitor, simultaneously collects a multi-source monitoring dataset containing raw pressure waveform data, instantaneous flow velocity data, and electrocardiogram synchronization signal data.

[0007] The original pressure waveform data is subjected to a four-level multi-scale decomposition based on the Daubechies wavelet basis function. The high-frequency coefficients of the first level are marked as cardiac cycle scale waveform components, the high-frequency coefficients of the second level are removed as transition frequency band coefficients, the low-frequency coefficients of the third level are marked as respiratory cycle scale waveform components, and the low-frequency coefficients of the fourth level are marked as low-frequency trend scale waveform components, thereby generating a set of multi-scale waveform components.

[0008] Time-frequency domain feature extraction is performed on the cardiac cycle-scale waveform components, respiratory cycle-scale waveform components, and low-frequency trend-scale waveform components in the multi-scale waveform component set. The approximate entropy value of the cardiac cycle-scale waveform component is calculated to generate a nonlinear waveform complexity index. The frequency band energy proportion of the respiratory cycle-scale waveform component is calculated to generate a spectral energy distribution feature vector. The Pearson correlation coefficient between the low-frequency trend-scale waveform component and the instantaneous flow velocity data is calculated to generate a pressure-flow velocity coupling coefficient. Then, combined with the patient's basic information, a hydrodynamic impedance model of the downstream vascular bed of the catheter is constructed, and the impedance change index of the downstream vascular bed of the catheter is derived.

[0009] The nonlinear waveform complexity index, spectral energy distribution feature vector, pressure-flow velocity coupling coefficient, and downstream vascular bed impedance change index are used as observation sequences input into a Kalman filter. Dynamic baseline drift modeling is performed through a sliding time window to estimate and generate a patient-specific pressure baseline model. At the same time, an adaptive catheter occlusion risk benchmark threshold is set based on the 3σ confidence interval of the prediction error of the patient-specific pressure baseline model.

[0010] The nonlinear waveform complexity index, spectral energy distribution feature vector, pressure-flow velocity coupling coefficient, downstream vascular bed impedance change index, and catheter occlusion risk benchmark threshold are input into a pre-defined Bayesian network. Based on a causal discovery algorithm, the network node connections are dynamically adjusted to construct a patient-specific risk factor graph structure. Causal inference calculations are then performed in conjunction with a patient-specific pressure baseline model to output a catheter occlusion probability index. A recurrent neural network is used to perform time-series prediction analysis on the feature sequence to generate a future risk evolution trend index. Finally, a hierarchical mapping process is performed on the catheter occlusion probability index and the future risk evolution trend index to generate a three-level warning signal including mild occlusion risk, moderate occlusion risk, and severe occlusion risk.

[0011] Preferably, the hemodynamic multimodal sensing system, which integrates a distal pressure sensor of the indwelling catheter, a flow sensor of the catheter line, and a patient's electrocardiogram monitor, simultaneously acquires a multi-source monitoring dataset containing raw pressure waveform data, instantaneous flow velocity data, and electrocardiogram synchronization signal data, including:

[0012] A multi-sensor spatiotemporal registration architecture was established, and the sampling frequency parameters of pressure, flow rate and electrocardiogram signals were set. The IEEE 1588 precise time protocol was used to achieve nanosecond-level timestamp alignment and generate original heterogeneous data packets with a unified time base. The original heterogeneous data packets include pressure sampling value sequences, flow rate sampling value sequences and electrocardiogram R wave marker sequences aligned according to the acquisition time.

[0013] Dynamic zero-point compensation of the catheter-blood coupling interface is performed on the original heterogeneous data package. The changes in blood viscosity coefficient and catheter elastic modulus are estimated in real time using an online calibration algorithm. A pressure baseline drift compensation model is established, and baseline drift is eliminated through the pressure baseline drift compensation model to generate drift-compensated pressure data.

[0014] Based on the drift-compensated pressure data, the velocity sampling sequence in the original heterogeneous data package, and the ECG R-wave marker sequence, vascular hydrostatic pressure geometric correction and hemodynamic dimension normalization are performed to generate a raw hemodynamic digital twin dataset. The raw hemodynamic digital twin dataset is stored in a circular buffer queue structure and an alignment index is established according to the timestamp.

[0015] Preferably, the original pressure waveform data undergoes a four-level multi-scale decomposition process based on the Daubechies wavelet basis function. The first-level high-frequency coefficients are labeled as cardiac cycle-scale waveform components, the second-level high-frequency coefficients are removed as transition band coefficients, the third-level low-frequency coefficients are labeled as respiratory cycle-scale waveform components, and the fourth-level low-frequency coefficients are labeled as low-frequency trend-scale waveform components, thereby generating a multi-scale waveform component set, including:

[0016] A multi-scale decomposition framework for catheter pressure waveforms was constructed. The db8 wavelet basis function was used to perform discrete wavelet transform on the pressure sample value sequence. The number of decomposition layers was set to 4. After each layer of decomposition, approximation coefficients and detail coefficients were obtained, generating a four-layer wavelet coefficient set containing the approximation coefficients and detail coefficients of each layer.

[0017] Physiological rhythm scale separation and noise immunity layer stripping are performed on the four-layer wavelet coefficient set. The first-layer detail coefficients are reconstructed into cardiac cycle scale waveform components, whose frequency bands correspond to the cardiac cycle. The third-layer approximation coefficients and the third-layer detail coefficients are jointly reconstructed into respiratory cycle scale waveform components, whose frequency bands correspond to the respiratory cycle. The fourth-layer approximation coefficients and the fourth-layer detail coefficients are jointly reconstructed into low-frequency trend scale waveform components, whose frequency bands correspond to the slow variation trend of vascular compliance. The second-layer detail coefficients are discarded because they correspond to the mechanical resonance interference layer. The first-layer approximation coefficients and the second-layer approximation coefficients are used as intermediate transit factors in the hierarchical decomposition and do not directly participate in the final reconstruction, generating a multi-scale waveform component set containing three physiologically relevant scales.

[0018] Preferably, the process involves performing time-frequency domain feature extraction on the cardiac cycle-scale waveform components, respiratory cycle-scale waveform components, and low-frequency trend-scale waveform components in the multi-scale waveform component set; calculating the approximate entropy value of the cardiac cycle-scale waveform components to generate a nonlinear waveform complexity index; calculating the frequency band energy proportion of the respiratory cycle-scale waveform components to generate a spectral energy distribution feature vector; and calculating the Pearson correlation coefficient between the low-frequency trend-scale waveform components and instantaneous flow velocity data to generate a pressure-flow velocity coupling coefficient. Furthermore, by combining this with the patient's basic information, a downstream vascular bed hydrodynamic impedance model is constructed, and a downstream vascular bed impedance change index is derived, including:

[0019] The nonlinear waveform complexity index is calculated based on the cardiac cycle-scale waveform components. The phase space of the cardiac cycle-scale waveform components is reconstructed, the embedding dimension m is 2, the time delay τ is 1, and the approximate entropy formula is used to generate the nonlinear waveform complexity index.

[0020] The spectral energy distribution feature vector is calculated based on the waveform components at the respiratory cycle scale. A fast Fourier transform is performed on the spectral energy distribution feature vector to obtain the power spectrum. The ratio of respiratory band energy to full-band energy is calculated to generate the spectral energy distribution feature vector.

[0021] The pressure-velocity coupling coefficient is calculated based on the low-frequency trend scale waveform component and the velocity sampling value sequence in the original digital twin dataset of hemodynamics. The two sequences are aligned in the time dimension, and the temporal correlation is calculated using the Pearson correlation coefficient formula to generate the pressure-velocity coupling coefficient.

[0022] Based on the nonlinear waveform complexity index, spectral energy distribution feature vector, pressure-flow velocity coupling coefficient, and age and disease type codes from the patient's basic information, a hydrodynamic impedance model of the downstream vascular bed was constructed. The model parameters were obtained through regression calibration using clinical data, and the impedance change index of the downstream vascular bed was calculated and generated.

[0023] Preferably, the nonlinear waveform complexity index, spectral energy distribution feature vector, pressure-flow velocity coupling coefficient, and downstream vascular bed impedance change index are used as observation sequences input into a Kalman filter. Dynamic baseline drift modeling is performed through a sliding time window to estimate and generate a patient-specific pressure baseline model. Simultaneously, based on the 3σ confidence interval of the prediction error of the patient-specific pressure baseline model, an adaptive catheter occlusion risk benchmark threshold is set, including:

[0024] Using the nonlinear waveform complexity index, spectral energy distribution feature vector, pressure-flow velocity coupling coefficient, and downstream vascular bed impedance change index as modeling inputs, a Kalman filter state space model is constructed. The state vector includes baseline drift, respiratory coupling coefficient drift, impedance coefficient drift, and dynamic impedance proportionality coefficient. The state transition equation describes the evolution of the state vector over time. The observation vector includes the nonlinear waveform complexity index, spectral energy distribution feature vector, pressure-flow velocity coupling coefficient, and downstream vascular bed impedance change index. The observation equation describes the linear mapping relationship between the observation vector and the state vector, thus generating the Kalman filter state space model structure.

[0025] A sliding window Bayesian parameter estimation was performed using a Kalman filter state-space model structure. The time window length was set to 120 seconds and the step size to 30 seconds. Kalman recursive update was performed in each window to calculate the optimal Kalman gain matrix and the state estimation sequence.

[0026] Based on the optimal Kalman gain matrix and state estimation sequence, the time series of baseline drift is extracted and its standard deviation is calculated. The occlusion risk benchmark threshold is set as the initial baseline mean plus or minus three times the standard deviation, thereby generating a patient-individualized pressure baseline model and an adaptive catheter occlusion risk benchmark threshold.

[0027] Preferably, the nonlinear waveform complexity index, spectral energy distribution feature vector, pressure-flow velocity coupling coefficient, downstream vascular bed impedance change index, and catheter occlusion risk benchmark threshold are input into a preset Bayesian network. Based on a causal discovery algorithm, the network node connections are dynamically adjusted to construct a patient-specific risk factor graph structure. This is then combined with a patient-specific pressure baseline model to perform causal inference calculations, outputting a catheter occlusion probability index. Furthermore, a recurrent neural network is used to perform time-series predictive analysis on the feature sequences to generate a future risk evolution trend index, including:

[0028] Based on a patient-specific pressure baseline model and an adaptive catheter occlusion risk benchmark threshold, a causal discovery algorithm is used to dynamically adjust the connection relationships of Bayesian network nodes. The nodes include a nonlinear waveform complexity index, a spectral energy distribution feature vector, a pressure-flow velocity coupling coefficient, a downstream vascular bed impedance change index, and a preset catheter occlusion risk latent variable. The edge structure is dynamically determined through a conditional independence test. The adaptive catheter occlusion risk benchmark threshold is used as a network hyperparameter to constrain the prior probability distribution of nodes, generating a dynamically adjusted patient-specific risk factor graph structure.

[0029] The nonlinear waveform complexity index, spectral energy distribution feature vector, pressure-flow velocity coupling coefficient, and downstream vascular bed impedance change index are input into the patient-specific risk factor graph structure. Bayesian causal inference is performed to calculate and output the catheter blockage probability index. At the same time, a recurrent neural network is used to perform time-series prediction analysis on the feature sequence. The time-series dependency is captured through the long short-term memory network structure to generate a future risk evolution trend index.

[0030] Preferably, the final step involves performing a hierarchical mapping process on the catheter blockage probability index and the future risk evolution trend index to generate a three-level early warning signal, including mild blockage risk, moderate blockage risk, and severe blockage risk, which includes:

[0031] A tiered mapping process is performed based on the catheter occlusion probability index and the future risk evolution trend index. The probability thresholds are set to 0.6 for level one and 0.8 for level two. When the catheter occlusion probability index is less than the level one threshold, a mild occlusion risk signal is generated. When the catheter occlusion probability index is greater than or equal to the level one threshold and less than the level two threshold, a moderate occlusion risk signal is generated. When the catheter occlusion probability index is greater than or equal to the level two threshold, a severe occlusion risk signal is generated. The future risk evolution trend index is used to predict the risk level change within a preset time period. The three-level warning signal is used to drive the audible and visual alarm device of the clinical monitoring terminal and trigger the automatic recording of electronic medical records.

[0032] Secondly, this application also provides an indwelling catheter blockage early warning system based on pressure waveform analysis, comprising:

[0033] Acquisition module: Used to simultaneously acquire multi-source monitoring datasets including raw pressure waveform data, instantaneous flow velocity data, and ECG synchronization signal data through a hemodynamic multimodal sensing system integrating a distal pressure sensor of the indwelling catheter, a flow sensor of the catheter line, and a patient ECG monitor;

[0034] Decomposition module: Used to perform four-level multi-scale decomposition processing on the original pressure waveform data based on the Daubechies wavelet basis function. The first-level high-frequency coefficients are marked as cardiac cycle scale waveform components, the second-level high-frequency coefficients are removed as transition frequency band coefficients, the third-level low-frequency coefficients are marked as respiratory cycle scale waveform components, and the fourth-level low-frequency coefficients are marked as low-frequency trend scale waveform components, thereby generating a set of multi-scale waveform components.

[0035] The computation module performs time-frequency domain feature extraction on the cardiac cycle-scale waveform components, respiratory cycle-scale waveform components, and low-frequency trend-scale waveform components in the multi-scale waveform component set. It calculates the approximate entropy value of the cardiac cycle-scale waveform components to generate a nonlinear waveform complexity index, calculates the frequency band energy proportion of the respiratory cycle-scale waveform components to generate a spectral energy distribution feature vector, and calculates the Pearson correlation coefficient between the low-frequency trend-scale waveform components and instantaneous flow velocity data to generate a pressure-flow velocity coupling coefficient. Furthermore, it constructs a hydrodynamic impedance model of the downstream vascular bed of the catheter by combining the patient's basic information and derives the impedance change index of the downstream vascular bed of the catheter.

[0036] Input estimation module: It is used to input the nonlinear waveform complexity index, spectral energy distribution feature vector, pressure-flow velocity coupling coefficient and downstream vascular bed impedance change index as observation sequence into Kalman filter, perform dynamic baseline drift modeling through sliding time window, estimate and generate patient-specific pressure baseline model, and set adaptive catheter occlusion risk benchmark threshold based on the 3σ confidence interval of the prediction error of patient-specific pressure baseline model.

[0037] The generation module is used to input the nonlinear waveform complexity index, spectral energy distribution feature vector, pressure-flow velocity coupling coefficient, downstream vascular bed impedance change index, and catheter occlusion risk benchmark threshold into a preset Bayesian network. Based on the causal discovery algorithm, it dynamically adjusts the network node connection relationship, constructs a patient-specific risk factor graph structure, and performs causal inference calculation processing in conjunction with the patient's individualized pressure baseline model to output the catheter occlusion probability index. Through recurrent neural network, it performs time-series prediction analysis on the feature sequence to generate a future risk evolution trend index. Finally, it performs hierarchical mapping processing on the catheter occlusion probability index and the future risk evolution trend index to generate a three-level early warning signal including mild occlusion risk, moderate occlusion risk, and severe occlusion risk.

[0038] Thirdly, this application also provides an indwelling catheter blockage early warning device based on pressure waveform analysis, comprising:

[0039] Memory, used to store computer programs;

[0040] A processor is used to implement the steps of the indwelling catheter blockage early warning method based on pressure waveform analysis when executing the computer program.

[0041] Fourthly, this application also provides a readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-described method for early warning of obstruction of indwelling catheters based on pressure waveform analysis.

[0042] The beneficial effects of this invention are as follows:

[0043] This invention constructs a heterogeneous sampling array with three modalities of intravascular pressure, flow velocity, and electrocardiogram, and achieves nanosecond-level spatiotemporal registration using the IEEE 1588 precise time protocol. Combined with an online calibration algorithm to dynamically estimate changes in blood viscosity coefficient and catheter elastic modulus, a pressure baseline drift compensation model is established. This significantly improves the time synchronization accuracy and measurement accuracy of multi-source data, eliminates baseline drift caused by long-term implantation, and provides a high-quality raw digital twin dataset of hemodynamics for subsequent analysis.

[0044] This invention employs Daubechies wavelet basis functions to perform four-layer multi-scale time-frequency atomic decomposition on pressure waveforms. By separating physiological rhythm scales and stripping noise immunity layers, the mechanical resonance interference layer is discarded, and the cardiac cycle scale waveform components, respiratory cycle scale waveform components, and low-frequency trend scale waveform components are accurately extracted. This achieves decoupling of different physiological mechanisms in pressure signals and improves the signal-to-noise ratio and physiological specificity of feature extraction.

[0045] This invention generates a nonlinear waveform complexity index by calculating the approximate entropy of the waveform components at the cardiac cycle scale, quantifies the frequency band energy ratio of the waveform components at the respiratory cycle scale to generate a spectral energy distribution feature vector, constructs a pressure-flow velocity coupling coefficient using the Pearson correlation coefficient, and further integrates patient age and disease type codes to construct a hydrodynamic impedance model of the downstream vascular bed of the catheter, deriving the impedance change index of the downstream vascular bed of the catheter. This extends traditional single pressure monitoring to multimodal coupling analysis, enabling the capture of subtle hemodynamic changes in the early stage of catheter occlusion, and improving the sensitivity and specificity of early warning.

[0046] This invention uses a multi-feature index as the input of the observation sequence into a Kalman filter, performs dynamic baseline drift modeling through a sliding time window, estimates and generates a patient-specific pressure baseline model, and sets an adaptive catheter occlusion risk benchmark threshold based on the 3σ confidence interval of the prediction error. This achieves a leap from a fixed population threshold to an individual adaptive threshold, effectively reducing the false alarm rate and false negative rate caused by patient physiological differences.

[0047] This invention employs a causal discovery algorithm to dynamically adjust the connection relationships of Bayesian network nodes, constructing a patient-specific risk factor graph structure. The nonlinear waveform complexity index, spectral energy distribution feature vector, pressure-flow velocity coupling coefficient, and downstream vascular bed impedance change index are input into this graph structure to perform Bayesian causal inference, outputting a catheter occlusion probability index. Simultaneously, a recurrent neural network (LSN) structure is used in parallel to perform time-series prediction of the feature sequences, generating a future risk evolution trend index. This represents a technological upgrade from static risk assessment to dynamic prognostic prediction, providing clinicians with the risk level change trend over a predetermined timeframe.

[0048] This invention performs hierarchical mapping processing based on the catheter occlusion probability index, the future risk evolution trend index, and the adaptive catheter occlusion risk benchmark threshold to generate a three-level early warning signal including mild occlusion risk, moderate occlusion risk, and severe occlusion risk. This signal directly drives the audible and visual alarm device of the clinical monitoring terminal and triggers automatic recording of electronic medical records, realizing hierarchical early warning and automated response of occlusion risk, and providing precise intervention decision support for clinical medical staff.

[0049] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing embodiments of the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description, claims, and drawings. Attached Figure Description

[0050] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1 This is a schematic diagram of the indwelling catheter blockage early warning method based on pressure waveform analysis as described in an embodiment of the present invention;

[0052] Figure 2 This is a schematic diagram of the indwelling catheter blockage early warning system based on pressure waveform analysis as described in an embodiment of the present invention;

[0053] Figure 3 This is a schematic diagram of the indwelling catheter blockage early warning device based on pressure waveform analysis as described in an embodiment of the present invention.

[0054] In the diagram: 701, Acquisition module; 702, Decomposition module; 703, Calculation module; 704, Input estimation module; 705, Construction and generation module; 800, Indwelling catheter blockage early warning device based on pressure waveform analysis; 801, Processor; 802, Memory; 803, Multimedia component; 804, I / O interface; 805, Communication component. Detailed Implementation

[0055] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0056] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this invention, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0057] Example 1:

[0058] This embodiment provides a method for early warning of indwelling catheter blockage based on pressure waveform analysis.

[0059] See Figure 1 The figure shows that the method includes steps S100, S200, S300, S400 and S500.

[0060] S100, through a hemodynamic multimodal sensing system integrating a distal pressure sensor of the indwelling catheter, a flow sensor of the catheter line, and a patient's electrocardiogram monitor, simultaneously acquires a multi-source monitoring dataset containing raw pressure waveform data, instantaneous flow velocity data, and electrocardiogram synchronization signal data.

[0061] It is understood that step S100 includes S101, S102, and S103, wherein:

[0062] S101. Establish a multi-sensor spatiotemporal registration architecture, set the sampling frequency parameters of pressure, flow rate and ECG signals, use the IEEE1588 precise time protocol to achieve nanosecond-level timestamp alignment, and generate original heterogeneous data packets with a unified time reference. The original heterogeneous data packets include pressure sampling value sequences, flow rate sampling value sequences and ECG R-wave marker sequences aligned according to the acquisition time.

[0063] S102. Perform dynamic zero-point compensation at the catheter-blood coupling interface on the original heterogeneous data package. Use an online calibration algorithm to estimate the changes in blood viscosity coefficient and catheter elastic modulus in real time, establish a pressure baseline drift compensation model, and eliminate baseline drift through the pressure baseline drift compensation model to generate drift-compensated pressure data. The calculation formula of the pressure baseline drift compensation model is as follows:

[0064]

[0065] In the formula, The calculated baseline drift amount, The blood viscosity coefficient was calibrated experimentally. The blood viscosity coefficient is estimated in real time using an online calibration algorithm. This is the preset effective length of the catheter. The elastic deformation coefficient of the catheter was determined experimentally. The elastic modulus of the conduit material is estimated in real time using an online calibration algorithm. For the radial deformation of the duct;

[0066] S103. Based on the pressure data after drift compensation, the velocity sampling value sequence in the original heterogeneous data package, and the ECG R-wave label sequence, perform vascular hydrostatic pressure geometric correction and hemodynamic dimension normalization to generate a hemodynamic original digital twin dataset. The hemodynamic original digital twin dataset is stored in a circular buffer queue structure and an alignment index is established according to the timestamp.

[0067] It should be noted that this step first establishes the anatomical coordinate transformation relationship between the catheter tip and the right atrium based on the patient's preoperative imaging data using a three-dimensional spatial positioning mapping algorithm. Combined with real-time data from the position sensor, the vertical height difference h between the sensor and the catheter is calculated. A dynamic blood density estimate ρ (adjusted in real-time based on hematocrit) is used to eliminate the influence of hydrostatic pressure gradients. This correction can reach 8-12 mmHg for internal jugular vein catheterization, effectively avoiding baseline misjudgment caused by changes in body position. Subsequently, hierarchical dimensional normalization is implemented: pressure signals are standardized using z-scores for cardiac cycle, respiratory cycle, and slow-varying trend bands respectively; flow velocity signals are robustly standardized using Box-Cox transformation combined with median absolute deviation to preserve pulse characteristics; and ECG R-wave markers are converted into heart rate variability frequency domain indicators for physiological driving reference. The final constructed digital twin dataset is stored using a circular buffer queue structure, achieving zero-copy access through memory mapping, and employing a two-layer hashing mechanism with minute-level coarse indexes and second-level fine indexes. Compared to traditional file-based storage, this architecture improves real-time response capability by two orders of magnitude, providing a standardized data foundation for subsequent algorithms that is spatiotemporally aligned, dimensionally unified, and accessible at the millisecond level, significantly reducing feature overload and computational errors caused by data heterogeneity.

[0068] S200. Perform a four-level multi-scale decomposition process based on the Daubechies wavelet basis function on the original pressure waveform data. Mark the high-frequency coefficients of the first level as cardiac cycle scale waveform components, remove the high-frequency coefficients of the second level as transition band coefficients, mark the low-frequency coefficients of the third level as respiratory cycle scale waveform components, and mark the low-frequency coefficients of the fourth level as low-frequency trend scale waveform components, thereby generating a set of multi-scale waveform components.

[0069] It is understood that step S200 includes S201 and S202, wherein:

[0070] S201. Construct a multi-scale decomposition framework for catheter pressure waveforms. Use the db8 wavelet basis function to perform discrete wavelet transform on the pressure sample value sequence. Set the number of decomposition layers to 4. After each layer of decomposition, obtain the approximation coefficients and detail coefficients, and generate a four-layer wavelet coefficient set containing the approximation coefficients and detail coefficients of each layer.

[0071] It should be noted that the db8 wavelet basis function selected in this embodiment is effective in suppressing boundary effects and accurately capturing transient changes in catheter pressure signals due to its 8th-order vanishing moment and tight support characteristics. Compared with the traditional Fourier transform, it is more suitable for non-stationary physiological signals. The four-layer decomposition structure was verified by physiological rhythm matching: the first layer of detail coefficients captures the cardiac cycle scale (0.5-4Hz), the second layer of detail coefficients identifies mechanical resonance interference (4-8Hz), the third layer of coefficients extracts the respiratory cycle scale (0.1-0.5Hz), and the fourth layer of coefficients separates the vascular compliance slow variation component (<0.1Hz). The Mallat fast algorithm is used to achieve real-time processing with O(n) linear complexity, and the single decomposition time is <3ms at a 500Hz sampling rate, meeting the real-time requirements of bedside monitoring. This decomposition framework decouples the original pressure signal into independent components of different physiological mechanisms, avoiding frequency band aliasing. Compared with the endpoint effect and mode confusion problem of empirical mode decomposition, it significantly improves the physiological interpretability and computational stability of the decomposition results, providing a set of pure components with clear physiological meaning for subsequent feature extraction.

[0072] S202. Perform physiological rhythm scale separation and noise immunity layer stripping on the four-layer wavelet coefficient set. Reconstruct the first layer detail coefficients into cardiac cycle scale waveform components, whose frequency band corresponds to the cardiac cycle. Reconstruct the third layer approximation coefficients and the third layer detail coefficients together into respiratory cycle scale waveform components, whose frequency band corresponds to the respiratory cycle. Reconstruct the fourth layer approximation coefficients and the fourth layer detail coefficients together into low-frequency trend scale waveform components, whose frequency band corresponds to the slow change trend of vascular compliance. The second layer detail coefficients are discarded because they correspond to the mechanical resonance interference layer. The first layer approximation coefficients and the second layer approximation coefficients are used as intermediate transfer quantities in the hierarchical decomposition and do not directly participate in the final reconstruction, generating a multi-scale waveform component set containing three physiologically relevant scales.

[0073] It should be noted that a hierarchical storage structure is constructed based on a multi-scale waveform component set. The multi-scale waveform component set includes three sets of data: cardiac cycle scale waveform components, respiratory cycle scale waveform components, and low-frequency trend scale waveform components. The time resolution of each set of data is one-half, one-eighth, and one-sixteenth of the time resolution of the original data, respectively. The hierarchical storage structure is constructed by aligning the index with the timestamp.

[0074] S300 performs time-frequency domain feature extraction on the cardiac cycle-scale waveform components, respiratory cycle-scale waveform components, and low-frequency trend-scale waveform components in the multi-scale waveform component set. It calculates the approximate entropy value of the cardiac cycle-scale waveform components to generate a nonlinear waveform complexity index, calculates the frequency band energy proportion of the respiratory cycle-scale waveform components to generate a spectral energy distribution feature vector, and calculates the Pearson correlation coefficient between the low-frequency trend-scale waveform components and instantaneous flow velocity data to generate a pressure-flow velocity coupling coefficient. Then, it constructs a hydrodynamic impedance model of the downstream vascular bed of the catheter by combining the patient's basic information and derives the impedance change index of the downstream vascular bed of the catheter.

[0075] It is understood that step S300 includes S301, S302, and S303, wherein:

[0076] S301. Calculate the nonlinear waveform complexity index based on cardiac cycle-scale waveform components. Reconstruct the phase space of the cardiac cycle-scale waveform components, with an embedding dimension of m = 2 and a time delay of τ = 1. Use an approximate entropy formula to generate the nonlinear waveform complexity index, as shown in the following formula:

[0077]

[0078] In the formula, Represents the length of the waveform component sequence on the cardiac cycle scale. Represents the embedding dimension. Represents the i-th Similarity probability of vectors in phase space. Represents the i-th Similarity probability of vectors in phase space. Represents an approximate entropy value;

[0079] S302. Calculate the spectral energy distribution feature vector based on the waveform components of the respiratory cycle scale, perform a fast Fourier transform on the spectral energy distribution feature vector to obtain the power spectrum, calculate the ratio of respiratory frequency band energy to full frequency band energy, and generate the spectral energy distribution feature vector.

[0080] It should be noted that step S302 estimates the power spectral density using the Welch method, selecting the respiratory frequency band of 0.1-0.5 Hz as the numerator and the full-band Nyquist frequency as the denominator to calculate the energy ratio. This ratio quantifies the coupling strength between respiratory drive and catheter pressure fluctuations. In the early stages of catheter occlusion, the respiratory-pressure coupling weakens due to increased downstream resistance, and this ratio decreases 2-4 hours earlier than changes in absolute pressure. The spectral energy ratio has better statistical robustness and is unaffected by occasional fluctuations in respiratory rhythm. In practical applications, this eigenvector is a scalar exponent, requiring minimal memory for storage, facilitating real-time calculation and transmission.

[0081] S303. Calculate the pressure-flow velocity coupling coefficient based on the low-frequency trend scale waveform component and the velocity sampling value sequence in the original digital twin dataset of hemodynamics. Align the two sequences in the time dimension and use the Pearson correlation coefficient formula to calculate the temporal correlation to generate the pressure-flow velocity coupling coefficient.

[0082] It should be noted that this step calculates the correlation coefficient between the low-frequency trend component and flow velocity within a 120-second sliding window. The low-frequency trend reflects the slow-wave component of vascular compliance, while flow velocity reflects instantaneous blood flow. The degree of coupling between the two directly characterizes the vascular bed impedance. A robust estimation method is used for the correlation coefficient, assigning low weights to outliers to prevent interference from occasional blood flow shocks. When microthrombi occur in the catheter, increased downstream vascular resistance leads to the breakage of the pressure-flow velocity coupling, resulting in a decrease in the absolute value of the correlation coefficient, significantly earlier than changes in the pressure waveform morphology. This coefficient ranges from -1 to 1, with negative values ​​indicating inverse correlation and positive values ​​indicating positive correlation. The risk level is determined by the magnitude of the absolute value, not the sign.

[0083] S304. Based on the nonlinear waveform complexity index, spectral energy distribution feature vector, pressure-flow velocity coupling coefficient, and the age and disease type codes in the patient's basic information, a hydrodynamic impedance model of the downstream vascular bed of the catheter is constructed. The model parameters are obtained through regression calibration of clinical data, and the impedance change index of the downstream vascular bed of the catheter is calculated and generated.

[0084] It should be noted that the formula for calculating its impedance change index is as follows:

[0085]

[0086] In the formula, The index representing the change in impedance of the vascular bed downstream of the catheter. Represents the patient's individual baseline impedance. to The coefficient represents the value obtained through regression analysis of clinical data. It represents the complexity index of nonlinear waveforms. The respiratory band energy ratio represents the characteristic vector of spectral energy distribution. Represents the pressure-velocity coupling coefficient. Represents the patient's age. The code represents the disease type and generates an index of impedance change in the downstream vascular bed of the catheter.

[0087] It should be noted that this step constructs an impedance fusion model through multiple linear regression, integrating the aforementioned three characteristic indices with patient age and disease type codes (e.g., diabetes coded as 1, tumor coded as 2) using weighted aggregation. The weights γ1 to γ5 are obtained based on retrospective data from 500 clinical catheter occlusion events screened using Lasso regression, effectively avoiding overfitting. The patient's individual baseline impedance Z0 is the mean of the data from the first stable hour after catheter placement.

[0088] S400 inputs the nonlinear waveform complexity index, spectral energy distribution feature vector, pressure-flow velocity coupling coefficient, and downstream vascular bed impedance change index of the catheter as observation sequences into a Kalman filter. Dynamic baseline drift modeling is performed through a sliding time window to estimate and generate a patient-specific pressure baseline model. At the same time, an adaptive catheter occlusion risk benchmark threshold is set based on the 3σ confidence interval of the prediction error of the patient-specific pressure baseline model.

[0089] It is understood that in this step, S400 includes S401, S402, and S403, wherein:

[0090] S401. Using the nonlinear waveform complexity index, spectral energy distribution feature vector, pressure-flow velocity coupling coefficient, and downstream vascular bed impedance change index as modeling inputs, a Kalman filter state space model is constructed. The state vector includes baseline drift, respiratory coupling coefficient drift, impedance coefficient drift, and dynamic impedance proportionality coefficient. The state transition equation describes the evolution of the state vector over time. The observation vector includes the nonlinear waveform complexity index, spectral energy distribution feature vector, pressure-flow velocity coupling coefficient, and downstream vascular bed impedance change index. The observation equation describes the linear mapping relationship between the observation vector and the state vector, thereby generating the Kalman filter state space model structure.

[0091] S402. Using a Kalman filter state-space model structure, perform sliding window Bayesian parameter estimation. Set the time window length to 120 seconds and the step size to 30 seconds. Perform Kalman recursive updates within each window to calculate the optimal Kalman gain matrix and the state estimation sequence. The formula for calculating the Kalman gain is as follows:

[0092]

[0093] In the formula, The representation of the state prediction error covariance matrix, Represents the observation matrix. Represents the transpose of the observation matrix. Represents the observation noise covariance matrix. Represents the optimal Kalman gain matrix;

[0094] S403. Based on the optimal Kalman gain matrix and state estimation sequence, extract the time series of baseline drift and calculate its standard deviation; set the occlusion risk benchmark threshold to the initial baseline mean plus or minus three times the standard deviation, thereby generating a patient-specific pressure baseline model and an adaptive catheter occlusion risk benchmark threshold.

[0095] It should be noted that this step extracts the baseline drift component from the four-dimensional state estimation sequence. This component reflects the slow-changing trend of pressure at the catheter-vessel interface, and its time series smoothing filters out high-frequency physiological fluctuations. The standard deviation is calculated using an exponentially weighted moving average, giving higher weight to recent data and making the threshold more sensitive to recent state changes. The exponential decay factor λ is set to 0.95, corresponding to approximately 20 minutes of effective memory time. The 3σ principle is used to set the threshold, increasing the confidence level and controlling the false alarm rate below 0.3% under the assumption that the drift is approximately normally distributed. The initial baseline mean is the mean of the state estimation during the first stable hour after catheter insertion (hours 2-3), avoiding the unstable period in the early stages of insertion. Actual clinical data shows that this adaptive threshold reduces the false alarm rate compared to a fixed threshold, and automatically relaxes the threshold for elderly patients (>65 years old) and patients with coagulation disorders to avoid false alarms triggered by physiological fluctuations, reflecting the precision of personalized medicine. The generated baseline model is stored in the form of parameters (including a 4-dimensional state vector, a 4×4 state covariance matrix, and upper and lower bounds of the threshold), occupying only 256 bytes of memory, and can be efficiently maintained in embedded monitoring devices.

[0096] The S500 algorithm inputs the nonlinear waveform complexity index, spectral energy distribution feature vector, pressure-flow velocity coupling coefficient, downstream vascular bed impedance change index, and catheter occlusion risk benchmark threshold into a pre-defined Bayesian network. Based on a causal discovery algorithm, it dynamically adjusts the network node connections to construct a patient-specific risk factor graph structure. Combined with a patient-specific pressure baseline model, it performs causal inference calculations and outputs a catheter occlusion probability index. Through a recurrent neural network, it performs time-series prediction analysis on the feature sequence to generate a future risk evolution trend index. Finally, it performs hierarchical mapping processing on the catheter occlusion probability index and the future risk evolution trend index to generate a three-level warning signal including mild occlusion risk, moderate occlusion risk, and severe occlusion risk.

[0097] It is understood that in this step, S500 includes S501, S502, and S503, wherein:

[0098] S501. Based on the patient-individualized pressure baseline model and the adaptive catheter occlusion risk benchmark threshold, a causal discovery algorithm is used to dynamically adjust the connection relationship of Bayesian network nodes. The nodes include nonlinear waveform complexity index, spectral energy distribution feature vector, pressure-flow velocity coupling coefficient, catheter downstream vascular bed impedance change index, and preset catheter occlusion risk latent variables. The edge structure is dynamically determined through conditional independence test. The adaptive catheter occlusion risk benchmark threshold is used as a network hyperparameter to constrain the prior probability distribution of nodes, generating a dynamically adjusted patient-specific risk factor graph structure.

[0099] It should be noted that this step breaks through the limitations of traditional preset static network structures, employing PC or GES algorithms to automatically learn causal dependencies from current patient monitoring data, identifying the differences in occlusion causes among different patients. For example, cancer patients may primarily follow a causal path of downstream vascular bed impedance change index → ​​occlusion risk, while elderly patients may exhibit a chain-like causal structure of nonlinear waveform complexity index → ​​impedance change index → ​​occlusion risk. The dynamic adjustment cycle of the edge structure is set to 6 hours to balance computational overhead with the speed of patient state evolution. An adaptive threshold is introduced as a hyperparameter into the Bayesian Dirichlet prior distribution, giving samples near the threshold greater probability quality and avoiding decision boundary jitter caused by hard thresholds. Compared to a fixed network structure, this dynamic method improves inference accuracy by approximately 15 percentage points, especially for newly admitted patients (with insufficient data). By loading a population prior network through transfer learning, it quickly converges to an individual-specific structure within 24 hours, solving the cold start problem.

[0100] S502. Input the nonlinear waveform complexity index, spectral energy distribution feature vector, pressure-flow velocity coupling coefficient, and catheter downstream vascular bed impedance change index into the patient-specific risk factor graph structure, perform Bayesian causal inference to calculate and output the catheter blockage probability index, and simultaneously use a recurrent neural network to perform time-series prediction analysis on the feature sequence, capture time-series dependencies through a long short-term memory network structure, and generate a future risk evolution trend index.

[0101] It should be noted that the "catheter occlusion risk latent variable" is a latent variable node that is pre-defined as part of the network structure when constructing the Bayesian network. It is used to characterize the potential risk state of duct occlusion that cannot be directly observed. Its prior probability distribution is constrained by the adaptive duct occlusion risk benchmark threshold generated in the above steps as a hyperparameter.

[0102] Understandably, this step implements a dual prediction mechanism: causal inference provides an explanatory probability of congestion risk at the current moment, while time-series prediction provides a trend index of future risk evolution. Bayesian inference uses variational inference to approximate the posterior distribution, and Kalman filter state estimation is introduced as the initial mean of the variational distribution during the ELBO lower bound optimization process to accelerate convergence. The Long Short-Term Memory (LSTM) network adopts a bidirectional structure, with a sequence length of 96 time points (corresponding to 8 hours of data) and a hidden layer dimension of 128. An attention mechanism weights features from the most recent hour, allowing the model to focus on recent trends. The two models run in parallel and share a feature cache, reducing computational redundancy.

[0103] The formula for calculating Bayesian causal inference is as follows:

[0104]

[0105] In the formula, This represents the posterior causal probability of ductal obstruction given the current observational evidence. This represents the causal likelihood probability of observing characteristic indicators under the current congestion risk state. θ represents the prior probability of blockage causality based on prior knowledge of duct dynamics, and θ represents the set of network hidden variable parameters.

[0106] S503. Based on the catheter occlusion probability index and the future risk evolution trend index, a hierarchical mapping process is performed. The probability thresholds are set to 0.6 for level one and 0.8 for level two. When the catheter occlusion probability index is less than the level one threshold, a mild occlusion risk signal is generated. When the catheter occlusion probability index is greater than or equal to the level one threshold and less than the level two threshold, a moderate occlusion risk signal is generated. When the catheter occlusion probability index is greater than or equal to the level two threshold, a severe occlusion risk signal is generated. The future risk evolution trend index is used to predict the risk level change within a preset time period. The three-level early warning signal is used to drive the audible and visual alarm device of the clinical monitoring terminal and trigger the automatic recording of electronic medical records.

[0107] It should be noted that in this step, the current risk level is determined by the probability index, while the future trend index is used to adjust the alarm sensitivity. When the trend index shows a future rate of increase exceeding a threshold (e.g., an increase of 0.15 per hour), even if the current risk level is mild, the system automatically raises the alarm level to moderate, achieving "risk acceleration" perception. The three-level warning signals use different colored LED lights and sound frequencies: mild (yellow light + 1Hz low-frequency sound), moderate (orange light + 2Hz mid-frequency sound), and severe (red light + 4Hz high-frequency sound), conforming to ergonomic design principles and reducing the cognitive load on medical staff in noisy ICU environments. The electronic medical record not only includes the alarm level but also automatically appends the current feature index vector, state estimation sequence, and a snapshot of the risk factor graph structure.

[0108] In summary, this invention achieves synchronous acquisition and spatiotemporal registration of multi-source data by constructing a heterogeneous sampling array of intravascular pressure, flow velocity, and electrocardiogram (ECG) modes. It employs Daubechies wavelet basis functions to perform four-layer multi-scale time-frequency atomic decomposition of the pressure waveform, accurately separating the cardiac cycle, respiratory cycle, and slow-varying trend components of vascular compliance. Based on this, it extracts approximate entropy, spectral energy ratio, and pressure-flow velocity coupling coefficient, and integrates basic information such as patient age and disease type to construct a hydrodynamic impedance model of the downstream vascular bed. A Kalman filter sliding window dynamic baseline model is used to generate an individualized pressure baseline model and adaptive risk threshold. Furthermore, a causal discovery algorithm is introduced to dynamically adjust the Bayesian network structure, forming a patient-specific risk factor map. Combined with a recurrent neural network, it performs time-series prediction, outputting a future risk evolution trend index, which is ultimately mapped to a three-level warning signal (mild, moderate, and severe). This method represents a technological leap from static monitoring to dynamic prediction, from population thresholds to individual models, and from single pressure interpretation to multimodal coupling analysis, significantly improving the sensitivity and specificity of early warning of catheter occlusion.

[0109] Example 2:

[0110] like Figure 2 As shown, this embodiment provides an indwelling catheter blockage early warning system based on pressure waveform analysis. See [link to documentation]. Figure 2 The system includes:

[0111] Acquisition module 701: Used to simultaneously acquire a multi-source monitoring dataset containing raw pressure waveform data, instantaneous flow velocity data and ECG synchronization signal data through a hemodynamic multimodal sensing system integrating a distal pressure sensor of the indwelling catheter, a flow sensor of the catheter line and a patient ECG monitor;

[0112] Decomposition module 702: is used to perform four-level multi-scale decomposition processing on the original pressure waveform data based on the Daubechies wavelet basis function. The first-level high-frequency coefficients are marked as cardiac cycle scale waveform components, the second-level high-frequency coefficients are removed as transition frequency band coefficients, the third-level low-frequency coefficients are marked as respiratory cycle scale waveform components, and the fourth-level low-frequency coefficients are marked as low-frequency trend scale waveform components, thereby generating a set of multi-scale waveform components.

[0113] Calculation module 703: This module performs time-frequency domain feature extraction on the cardiac cycle-scale waveform components, respiratory cycle-scale waveform components, and low-frequency trend-scale waveform components in the multi-scale waveform component set. It calculates the approximate entropy value of the cardiac cycle-scale waveform components to generate a nonlinear waveform complexity index, calculates the frequency band energy proportion of the respiratory cycle-scale waveform components to generate a spectral energy distribution feature vector, and calculates the Pearson correlation coefficient between the low-frequency trend-scale waveform components and instantaneous flow velocity data to generate a pressure-flow velocity coupling coefficient. Furthermore, it combines the patient's basic information to construct a hydrodynamic impedance model of the downstream vascular bed and derives the impedance change index of the downstream vascular bed.

[0114] Input estimation module 704: is used to input the nonlinear waveform complexity index, spectral energy distribution feature vector, pressure-flow velocity coupling coefficient and downstream vascular bed impedance change index as observation sequence into the Kalman filter, perform dynamic baseline drift modeling through a sliding time window, estimate and generate a patient-specific pressure baseline model, and set an adaptive catheter occlusion risk benchmark threshold based on the 3σ confidence interval of the prediction error of the patient-specific pressure baseline model.

[0115] The generation module 705 is used to input the nonlinear waveform complexity index, spectral energy distribution feature vector, pressure-flow velocity coupling coefficient, downstream vascular bed impedance change index, and catheter occlusion risk benchmark threshold into a preset Bayesian network. Based on the causal discovery algorithm, it dynamically adjusts the network node connection relationship, constructs a patient-specific risk factor graph structure, and performs causal inference calculation processing in conjunction with the patient's individualized pressure baseline model to output the catheter occlusion probability index. Through recurrent neural network, it performs time-series prediction analysis on the feature sequence to generate a future risk evolution trend index. Finally, it performs hierarchical mapping processing on the catheter occlusion probability index and the future risk evolution trend index to generate a three-level early warning signal including mild occlusion risk, moderate occlusion risk, and severe occlusion risk.

[0116] Specifically, the acquisition module 701 includes:

[0117] The generation unit is set up to establish a multi-sensor spatiotemporal registration architecture, set the sampling frequency parameters of pressure, flow rate and ECG signals, use the IEEE 1588 precise time protocol to achieve nanosecond-level timestamp alignment, and generate original heterogeneous data packets with a unified time base. The original heterogeneous data packets include pressure sampling value sequences, flow rate sampling value sequences and ECG R-wave marker sequences aligned according to the acquisition time.

[0118] A model unit is established to perform dynamic zero-point compensation of the catheter-blood coupling interface on the original heterogeneous data package. An online calibration algorithm is used to estimate the changes in blood viscosity coefficient and catheter elastic modulus in real time. A pressure baseline drift compensation model is established, and baseline drift is eliminated using this model. Drift-compensated pressure data is then generated. The calculation formula for the pressure baseline drift compensation model is as follows:

[0119]

[0120] In the formula, The calculated baseline drift amount, The blood viscosity coefficient was calibrated experimentally. The blood viscosity coefficient is estimated in real time using an online calibration algorithm. This is the preset effective length of the catheter. The elastic deformation coefficient of the catheter was determined experimentally. The elastic modulus of the conduit material is estimated in real time using an online calibration algorithm. For the radial deformation of the duct;

[0121] Execution unit: Based on the drift-compensated pressure data, the velocity sampling value sequence in the original heterogeneous data package, and the ECG R-wave label sequence, it performs vascular hydrostatic pressure geometric correction and hemodynamic dimension normalization to generate a raw hemodynamic digital twin dataset. The raw hemodynamic digital twin dataset is stored in a circular buffer queue structure and an alignment index is established according to the timestamp.

[0122] Specifically, the decomposition module 702 includes:

[0123] Construction Unit: Used to construct a multi-scale decomposition framework for catheter pressure waveforms. The db8 wavelet basis function is used to perform discrete wavelet transform on the pressure sample value sequence. The number of decomposition layers is set to 4. After each layer of decomposition, approximation coefficients and detail coefficients are obtained, generating a four-layer wavelet coefficient set containing the approximation coefficients and detail coefficients of each layer.

[0124] The stripping and reconstruction unit is used to perform physiological rhythm scale separation and noise immunity layer stripping on the four-layer wavelet coefficient set. The first layer detail coefficients are reconstructed into cardiac cycle scale waveform components, whose frequency bands correspond to the cardiac cycle. The third layer approximation coefficients and the third layer detail coefficients are jointly reconstructed into respiratory cycle scale waveform components, whose frequency bands correspond to the respiratory cycle. The fourth layer approximation coefficients and the fourth layer detail coefficients are jointly reconstructed into low-frequency trend scale waveform components, whose frequency bands correspond to the slow variation trend of vascular compliance. The second layer detail coefficients are discarded because they correspond to the mechanical resonance interference layer. The first layer approximation coefficients and the second layer approximation coefficients are used as intermediate transit factors in the hierarchical decomposition and do not directly participate in the final reconstruction, generating a multi-scale waveform component set containing three physiologically relevant scales.

[0125] Specifically, the computing module 703 includes:

[0126] The first computational unit is used to calculate the nonlinear waveform complexity index based on the cardiac cycle-scale waveform components. It reconstructs the phase space of the cardiac cycle-scale waveform components, with an embedding dimension m of 2 and a time delay τ of 1. It then uses an approximate entropy formula to generate the nonlinear waveform complexity index, the calculation formula of which is as follows:

[0127]

[0128] In the formula, Represents the length of the waveform component sequence on the cardiac cycle scale. Represents the embedding dimension. Represents the i-th Similarity probability of vectors in phase space. Represents the i-th Similarity probability of vectors in phase space. Represents an approximate entropy value;

[0129] The second calculation unit is used to calculate the spectral energy distribution feature vector based on the waveform components of the respiratory cycle scale, perform a fast Fourier transform on the spectral energy distribution feature vector to obtain the power spectrum, calculate the ratio of respiratory band energy to full band energy, and generate the spectral energy distribution feature vector.

[0130] The third calculation unit is used to calculate the pressure-flow velocity coupling coefficient based on the low-frequency trend scale waveform component and the velocity sampling value sequence in the original digital twin dataset of hemodynamics. The two sequences are aligned in the time dimension, and the time correlation is calculated using the Pearson correlation coefficient formula to generate the pressure-flow velocity coupling coefficient.

[0131] The fourth calculation unit is used to construct a hydrodynamic impedance model of the downstream vascular bed of the catheter based on the nonlinear waveform complexity index, spectral energy distribution feature vector, pressure-flow velocity coupling coefficient, and age and disease type codes in the patient's basic information. The model parameters are obtained through clinical data regression calibration, and the impedance change index of the downstream vascular bed of the catheter is calculated and generated.

[0132] Specifically, the input estimation module 704 includes:

[0133] The generation unit is used to construct a Kalman filter state space model by taking the nonlinear waveform complexity index, spectral energy distribution feature vector, pressure-flow velocity coupling coefficient, and downstream vascular bed impedance change index as modeling inputs. The state vector includes baseline drift, respiratory coupling coefficient drift, impedance coefficient drift, and dynamic impedance proportionality coefficient. The state transition equation describes the evolution of the state vector over time. The observation vector includes the nonlinear waveform complexity index, spectral energy distribution feature vector, pressure-flow velocity coupling coefficient, and downstream vascular bed impedance change index. The observation equation describes the linear mapping relationship between the observation vector and the state vector, thereby generating the Kalman filter state space model structure.

[0134] The update unit is configured to perform sliding window Bayesian parameter estimation using a Kalman filter state-space model structure. The time window length is set to 120 seconds, with a step size of 30 seconds. Within each window, a Kalman recursive update is performed to calculate the optimal Kalman gain matrix and the state estimation sequence. The formula for calculating the Kalman gain is as follows:

[0135]

[0136] In the formula, The representation of the state prediction error covariance matrix, Represents the observation matrix. Represents the transpose of the observation matrix. Represents the observation noise covariance matrix. Represents the optimal Kalman gain matrix;

[0137] Extraction and calculation unit: used to extract the time series of baseline drift based on the optimal Kalman gain matrix and state estimation sequence, and calculate its standard deviation; set the occlusion risk benchmark threshold to the initial baseline mean plus or minus three times the standard deviation, thereby generating a patient-individualized pressure baseline model and an adaptive catheter occlusion risk benchmark threshold.

[0138] Specifically, the construction generation module 705 includes:

[0139] Dynamic adjustment unit: Based on the patient's individualized pressure baseline model and an adaptive catheter occlusion risk benchmark threshold, the unit uses a causal discovery algorithm to dynamically adjust the connection relationships of Bayesian network nodes. The nodes include nonlinear waveform complexity index, spectral energy distribution feature vector, pressure-flow velocity coupling coefficient, downstream vascular bed impedance change index, and preset catheter occlusion risk latent variables. The edge structure is dynamically determined through conditional independence test. The adaptive catheter occlusion risk benchmark threshold is used as a network hyperparameter to constrain the prior probability distribution of nodes, generating a dynamically adjusted patient-specific risk factor graph structure.

[0140] Execution Unit: This unit is used to input the nonlinear waveform complexity index, spectral energy distribution feature vector, pressure-flow velocity coupling coefficient, and downstream vascular bed impedance change index into the patient-specific risk factor graph structure, perform Bayesian causal inference to calculate and output the catheter blockage probability index, and simultaneously use a recurrent neural network to perform time-series prediction analysis on the feature sequence. By capturing time-series dependencies through a long short-term memory network structure, it generates a future risk evolution trend index.

[0141] Specifically, the construction and generation module 705 further includes:

[0142] Mapping prediction unit: Used to perform hierarchical mapping processing based on catheter occlusion probability index and future risk evolution trend index. The probability thresholds are set to 0.6 for level 1 and 0.8 for level 2. When the catheter occlusion probability index is less than the level 1 threshold, a mild occlusion risk signal is generated. When the catheter occlusion probability index is greater than or equal to the level 1 threshold and less than the level 2 threshold, a moderate occlusion risk signal is generated. When the catheter occlusion probability index is greater than or equal to the level 2 threshold, a severe occlusion risk signal is generated. The future risk evolution trend index is used to predict the risk level change within a preset time period. The three-level early warning signal is used to drive the audible and visual alarm device of the clinical monitoring terminal and trigger the automatic recording of electronic medical records.

[0143] It should be noted that the specific methods by which each module performs operations in the system described in the above embodiments have been described in detail in the embodiments related to the method, and will not be elaborated here.

[0144] Example 3:

[0145] Corresponding to the above method embodiments, this embodiment also provides an indwelling catheter blockage early warning device based on pressure waveform analysis. The indwelling catheter blockage early warning device based on pressure waveform analysis described below and the indwelling catheter blockage early warning method based on pressure waveform analysis described above can be referred to in correspondence.

[0146] Figure 3 This is a block diagram illustrating an indwelling catheter blockage early warning device 800 based on pressure waveform analysis, according to an exemplary embodiment. Figure 3 As shown, the indwelling catheter blockage early warning device 800 based on pressure waveform analysis includes a processor 801 and a memory 802. The indwelling catheter blockage early warning device 800 also includes one or more of a multimedia component 803, an I / O interface 804, and a communication component 805.

[0147] The processor 801 controls the overall operation of the indwelling catheter blockage early warning device 800 based on pressure waveform analysis to complete all or part of the steps in the aforementioned indwelling catheter blockage early warning method based on pressure waveform analysis. The memory 802 stores various types of data to support the operation of the indwelling catheter blockage early warning device 800. This data may include, for example, instructions for any application or method operating on the indwelling catheter blockage early warning device 800, as well as application-related data such as contact data, sent and received messages, images, audio, video, etc. The memory 802 can be implemented using any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The multimedia component 803 may include a screen and an audio component. The screen may be, for example, a touchscreen, and the audio component is used to output and / or input audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signals may be further stored in the memory 802 or transmitted via the communication component 805. The audio component also includes at least one speaker for outputting audio signals. I / O interface 804 provides an interface between processor 801 and other interface modules, such as a keyboard, mouse, or buttons. These buttons can be virtual or physical. Communication component 805 is used for wired or wireless communication between the pressure waveform analysis-based indwelling catheter blockage early warning device 800 and other devices. Wireless communication includes, for example, Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, or 4G, or a combination thereof. Therefore, the corresponding communication component 805 may include a Wi-Fi module, a Bluetooth module, or an NFC module.

[0148] In an exemplary embodiment, the indwelling catheter blockage early warning device 800 based on pressure waveform analysis may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the aforementioned indwelling catheter blockage early warning method based on pressure waveform analysis.

[0149] In another exemplary embodiment, a computer-readable storage medium including program instructions is also provided, which, when executed by a processor, implement the steps of the aforementioned indwelling catheter blockage early warning method based on pressure waveform analysis. For example, the computer-readable storage medium may be the aforementioned memory 802 including program instructions, which may be executed by the processor 801 of the indwelling catheter blockage early warning device 800 based on pressure waveform analysis to complete the aforementioned indwelling catheter blockage early warning method based on pressure waveform analysis.

[0150] Example 4:

[0151] Corresponding to the above method embodiments, this embodiment also provides a readable storage medium. The readable storage medium described below can be referred to in conjunction with the indwelling catheter blockage early warning method based on pressure waveform analysis described above.

[0152] A computer program is stored on a readable storage medium, and when the computer program is executed by a processor, it implements the steps of the indwelling catheter blockage early warning method based on pressure waveform analysis in the above method embodiments.

[0153] Specifically, the readable storage medium can be a USB flash drive, a portable hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, or any other readable storage medium capable of storing program code.

[0154] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for early warning of indwelling catheter blockage based on pressure waveform analysis, characterized in that, include: The hemodynamic multimodal sensing system, which integrates a distal pressure sensor of the indwelling catheter, a flow sensor of the catheter line, and a patient's electrocardiogram monitor, simultaneously collects a multi-source monitoring dataset containing raw pressure waveform data, instantaneous flow velocity data, and electrocardiogram synchronization signal data. The original pressure waveform data is subjected to a four-level multi-scale decomposition based on the Daubechies wavelet basis function. The high-frequency coefficients of the first level are marked as cardiac cycle scale waveform components, the high-frequency coefficients of the second level are removed as transition frequency band coefficients, the low-frequency coefficients of the third level are marked as respiratory cycle scale waveform components, and the low-frequency coefficients of the fourth level are marked as low-frequency trend scale waveform components, thereby generating a set of multi-scale waveform components. Time-frequency domain feature extraction is performed on the cardiac cycle-scale waveform components, respiratory cycle-scale waveform components, and low-frequency trend-scale waveform components in the multi-scale waveform component set. The approximate entropy value of the cardiac cycle-scale waveform component is calculated to generate a nonlinear waveform complexity index. The frequency band energy proportion of the respiratory cycle-scale waveform component is calculated to generate a spectral energy distribution feature vector. The Pearson correlation coefficient between the low-frequency trend-scale waveform component and the instantaneous flow velocity data is calculated to generate a pressure-flow velocity coupling coefficient. Then, combined with the patient's basic information, a hydrodynamic impedance model of the downstream vascular bed of the catheter is constructed, and the impedance change index of the downstream vascular bed of the catheter is derived. The nonlinear waveform complexity index, spectral energy distribution feature vector, pressure-flow velocity coupling coefficient, and downstream vascular bed impedance change index are used as observation sequences input into a Kalman filter. Dynamic baseline drift modeling is performed through a sliding time window to estimate and generate a patient-specific pressure baseline model. At the same time, an adaptive catheter occlusion risk benchmark threshold is set based on the 3σ confidence interval of the prediction error of the patient-specific pressure baseline model. The nonlinear waveform complexity index, spectral energy distribution feature vector, pressure-flow velocity coupling coefficient, downstream vascular bed impedance change index, and catheter occlusion risk benchmark threshold are input into a pre-defined Bayesian network. Based on a causal discovery algorithm, the network node connections are dynamically adjusted to construct a patient-specific risk factor graph structure. Causal inference calculations are then performed in conjunction with a patient-specific pressure baseline model to output a catheter occlusion probability index. A recurrent neural network is used to perform time-series prediction analysis on the feature sequence to generate a future risk evolution trend index. Finally, a hierarchical mapping process is performed on the catheter occlusion probability index and the future risk evolution trend index to generate a three-level warning signal including mild occlusion risk, moderate occlusion risk, and severe occlusion risk.

2. The method for early warning of indwelling catheter blockage based on pressure waveform analysis according to claim 1, characterized in that, The hemodynamic multimodal sensing system, integrating a distal pressure sensor of the indwelling catheter, a flow sensor in the catheter line, and a patient's electrocardiogram monitor, simultaneously acquires a multi-source monitoring dataset containing raw pressure waveform data, instantaneous flow velocity data, and electrocardiogram synchronization signal data, including: A multi-sensor spatiotemporal registration architecture was established, and the sampling frequency parameters of pressure, flow rate and electrocardiogram signals were set. The IEEE 1588 precise time protocol was used to achieve nanosecond-level timestamp alignment and generate original heterogeneous data packets with a unified time base. The original heterogeneous data packets include pressure sampling value sequences, flow rate sampling value sequences and electrocardiogram R wave marker sequences aligned according to the acquisition time. Dynamic zero-point compensation of the catheter-blood coupling interface was performed on the original heterogeneous data packets. An online calibration algorithm was used to estimate the changes in blood viscosity coefficient and catheter elastic modulus in real time, and a pressure baseline drift compensation model was established. The baseline drift was eliminated by the pressure baseline drift compensation model, and drift-compensated pressure data was generated. The calculation formula of the pressure baseline drift compensation model is as follows: In the formula, The calculated baseline drift amount, The blood viscosity coefficient was calibrated experimentally. The blood viscosity coefficient is estimated in real time using an online calibration algorithm. This is the preset effective length of the catheter. The elastic deformation coefficient of the catheter was determined experimentally. The elastic modulus of the conduit material is estimated in real time using an online calibration algorithm. For the radial deformation of the duct; Based on the drift-compensated pressure data, the velocity sampling sequence in the original heterogeneous data package, and the ECG R-wave marker sequence, vascular hydrostatic pressure geometric correction and hemodynamic dimension normalization are performed to generate a raw hemodynamic digital twin dataset. The raw hemodynamic digital twin dataset is stored in a circular buffer queue structure and an alignment index is established according to the timestamp.

3. The method for early warning of indwelling catheter blockage based on pressure waveform analysis according to claim 2, characterized in that, The original pressure waveform data is subjected to a four-level multi-scale decomposition process based on the Daubechies wavelet basis function. The first-level high-frequency coefficients are marked as cardiac cycle-scale waveform components, the second-level high-frequency coefficients are removed as transition band coefficients, the third-level low-frequency coefficients are marked as respiratory cycle-scale waveform components, and the fourth-level low-frequency coefficients are marked as low-frequency trend-scale waveform components, thereby generating a multi-scale waveform component set, including: A multi-scale decomposition framework for catheter pressure waveforms was constructed. The db8 wavelet basis function was used to perform discrete wavelet transform on the pressure sample value sequence. The number of decomposition layers was set to 4. After each layer of decomposition, approximation coefficients and detail coefficients were obtained, generating a four-layer wavelet coefficient set containing the approximation coefficients and detail coefficients of each layer. Physiological rhythm scale separation and noise immunity layer stripping are performed on the four-layer wavelet coefficient set. The first-layer detail coefficients are reconstructed into cardiac cycle scale waveform components, whose frequency bands correspond to the cardiac cycle. The third-layer approximation coefficients and the third-layer detail coefficients are jointly reconstructed into respiratory cycle scale waveform components, whose frequency bands correspond to the respiratory cycle. The fourth-layer approximation coefficients and the fourth-layer detail coefficients are jointly reconstructed into low-frequency trend scale waveform components, whose frequency bands correspond to the slow variation trend of vascular compliance. The second-layer detail coefficients are discarded because they correspond to the mechanical resonance interference layer. The first-layer approximation coefficients and the second-layer approximation coefficients are used as intermediate transit factors in the hierarchical decomposition and do not directly participate in the final reconstruction, generating a multi-scale waveform component set containing three physiologically relevant scales.

4. The method for early warning of indwelling catheter blockage based on pressure waveform analysis according to claim 1, characterized in that, The process involves performing time-frequency domain feature extraction on the cardiac cycle-scale waveform components, respiratory cycle-scale waveform components, and low-frequency trend-scale waveform components in the multi-scale waveform component set. The approximate entropy value of the cardiac cycle-scale waveform component is calculated to generate a nonlinear waveform complexity index. The frequency band energy proportion of the respiratory cycle-scale waveform component is calculated to generate a spectral energy distribution feature vector. The Pearson correlation coefficient between the low-frequency trend-scale waveform component and the instantaneous flow velocity data is calculated to generate a pressure-flow velocity coupling coefficient. Furthermore, by combining the patient's basic information, a hydrodynamic impedance model of the downstream vascular bed is constructed, and an impedance change index of the downstream vascular bed is derived, including: The nonlinear waveform complexity index is calculated based on the cardiac cycle-scale waveform components. Phase space reconstruction is performed on the cardiac cycle-scale waveform components, with an embedding dimension m of 2 and a time delay τ of 1. An approximate entropy formula is used to generate the nonlinear waveform complexity index, as shown in the following formula: In the formula, Represents the length of the waveform component sequence on the cardiac cycle scale. Represents the embedding dimension. Represents the i-th Similarity probability of vectors in phase space. Represents the i-th Similarity probability of vectors in phase space. Represents an approximate entropy value; The spectral energy distribution feature vector is calculated based on the waveform components at the respiratory cycle scale. A fast Fourier transform is performed on the spectral energy distribution feature vector to obtain the power spectrum. The ratio of respiratory band energy to full-band energy is calculated to generate the spectral energy distribution feature vector. The pressure-velocity coupling coefficient is calculated based on the low-frequency trend scale waveform component and the velocity sampling value sequence in the original digital twin dataset of hemodynamics. The two sequences are aligned in the time dimension, and the temporal correlation is calculated using the Pearson correlation coefficient formula to generate the pressure-velocity coupling coefficient. Based on the nonlinear waveform complexity index, spectral energy distribution feature vector, pressure-flow velocity coupling coefficient, and the age and disease type codes in the patient's basic information, a hydrodynamic impedance model of the downstream vascular bed of the catheter is constructed. The model parameters are obtained through regression calibration of clinical data, and the impedance change index of the downstream vascular bed of the catheter is calculated and generated.

5. The method for early warning of indwelling catheter blockage based on pressure waveform analysis according to claim 1, characterized in that, The nonlinear waveform complexity index, spectral energy distribution feature vector, pressure-flow velocity coupling coefficient, and downstream vascular bed impedance change index are used as observation sequences input into a Kalman filter. Dynamic baseline drift modeling is performed through a sliding time window to estimate and generate a patient-specific pressure baseline model. Simultaneously, based on the 3σ confidence interval of the prediction error of the patient-specific pressure baseline model, an adaptive catheter occlusion risk benchmark threshold is set, including: Using the nonlinear waveform complexity index, spectral energy distribution feature vector, pressure-flow velocity coupling coefficient, and downstream vascular bed impedance change index as modeling inputs, a Kalman filter state space model is constructed. The state vector includes baseline drift, respiratory coupling coefficient drift, impedance coefficient drift, and dynamic impedance proportionality coefficient. The state transition equation describes the evolution of the state vector over time. The observation vector includes the nonlinear waveform complexity index, spectral energy distribution feature vector, pressure-flow velocity coupling coefficient, and downstream vascular bed impedance change index. The observation equation describes the linear mapping relationship between the observation vector and the state vector, thus generating the Kalman filter state space model structure. A sliding window Bayesian parameter estimation is performed using a Kalman filter state-space model structure. The time window length is set to 120 seconds and the step size to 30 seconds. Kalman recursive updates are performed within each window to calculate the optimal Kalman gain matrix and the state estimation sequence. The formula for calculating the Kalman gain is as follows: In the formula, The representation of the state prediction error covariance matrix, Represents the observation matrix. Represents the transpose of the observation matrix. Represents the observation noise covariance matrix. Represents the optimal Kalman gain matrix; Based on the optimal Kalman gain matrix and state estimation sequence, the time series of baseline drift is extracted and its standard deviation is calculated. The occlusion risk benchmark threshold is set as the initial baseline mean plus or minus three times the standard deviation, thereby generating a patient-individualized pressure baseline model and an adaptive catheter occlusion risk benchmark threshold.

6. The method for early warning of indwelling catheter blockage based on pressure waveform analysis according to claim 1, characterized in that, The nonlinear waveform complexity index, spectral energy distribution feature vector, pressure-flow velocity coupling coefficient, downstream vascular bed impedance change index, and catheter blockage risk benchmark threshold are input into a preset Bayesian network. Based on the causal discovery algorithm, the network node connection relationship is dynamically adjusted to construct a patient-specific risk factor graph structure. Then, causal inference calculation is performed in combination with the patient's individualized pressure baseline model to output the catheter blockage probability index. By performing time-series predictive analysis on feature sequences using recurrent neural networks, a future risk evolution trend index is generated, which includes: Based on a patient-specific pressure baseline model and an adaptive catheter occlusion risk benchmark threshold, a causal discovery algorithm is used to dynamically adjust the connection relationships of Bayesian network nodes. The nodes include a nonlinear waveform complexity index, a spectral energy distribution feature vector, a pressure-flow velocity coupling coefficient, a downstream vascular bed impedance change index, and a preset catheter occlusion risk latent variable. The edge structure is dynamically determined through a conditional independence test. The adaptive catheter occlusion risk benchmark threshold is used as a network hyperparameter to constrain the prior probability distribution of nodes, generating a dynamically adjusted patient-specific risk factor graph structure. The nonlinear waveform complexity index, spectral energy distribution feature vector, pressure-flow velocity coupling coefficient, and downstream vascular bed impedance change index are input into the patient-specific risk factor graph structure. Bayesian causal inference is performed to calculate and output the catheter blockage probability index. At the same time, a recurrent neural network is used to perform time-series prediction analysis on the feature sequence. The time-series dependency is captured through the long short-term memory network structure to generate a future risk evolution trend index.

7. The method for early warning of indwelling catheter blockage based on pressure waveform analysis according to claim 6, characterized in that, The calculation formula for Bayesian causal inference is as follows: In the formula, This represents the posterior causal probability of ductal obstruction given the current observational evidence. This represents the causal likelihood probability of observing characteristic indicators under the current congestion risk state. θ represents the prior probability of blockage causality based on prior knowledge of duct dynamics, and θ represents the set of network hidden variable parameters.

8. The method for early warning of indwelling catheter blockage based on pressure waveform analysis according to claim 1, characterized in that, The final step involves performing a hierarchical mapping process on the catheter blockage probability index and the future risk evolution trend index to generate a three-level early warning signal, including mild blockage risk, moderate blockage risk, and severe blockage risk. A tiered mapping process is performed based on the catheter occlusion probability index and the future risk evolution trend index. The probability thresholds are set to 0.6 for level one and 0.8 for level two. When the catheter occlusion probability index is less than the level one threshold, a mild occlusion risk signal is generated. When the catheter occlusion probability index is greater than or equal to the level one threshold and less than the level two threshold, a moderate occlusion risk signal is generated. When the catheter occlusion probability index is greater than or equal to the level two threshold, a severe occlusion risk signal is generated. The future risk evolution trend index is used to predict the risk level change within a preset time period. The three-level warning signal is used to drive the audible and visual alarm device of the clinical monitoring terminal and trigger the automatic recording of electronic medical records.

9. A pre-emptive system for indwelling catheter blockage based on pressure waveform analysis, comprising the pre-emptive method for indwelling catheter blockage based on pressure waveform analysis as described in claim 1, characterized in that, include: Acquisition module: Used to simultaneously acquire multi-source monitoring datasets including raw pressure waveform data, instantaneous flow velocity data, and ECG synchronization signal data through a hemodynamic multimodal sensing system integrating a distal pressure sensor of the indwelling catheter, a flow sensor of the catheter line, and a patient ECG monitor; Decomposition module: Used to perform four-level multi-scale decomposition processing on the original pressure waveform data based on the Daubechies wavelet basis function. The first-level high-frequency coefficients are marked as cardiac cycle scale waveform components, the second-level high-frequency coefficients are removed as transition frequency band coefficients, the third-level low-frequency coefficients are marked as respiratory cycle scale waveform components, and the fourth-level low-frequency coefficients are marked as low-frequency trend scale waveform components, thereby generating a set of multi-scale waveform components. The computation module performs time-frequency domain feature extraction on the cardiac cycle-scale waveform components, respiratory cycle-scale waveform components, and low-frequency trend-scale waveform components in the multi-scale waveform component set. It calculates the approximate entropy value of the cardiac cycle-scale waveform components to generate a nonlinear waveform complexity index, calculates the frequency band energy proportion of the respiratory cycle-scale waveform components to generate a spectral energy distribution feature vector, and calculates the Pearson correlation coefficient between the low-frequency trend-scale waveform components and instantaneous flow velocity data to generate a pressure-flow velocity coupling coefficient. Furthermore, it constructs a hydrodynamic impedance model of the downstream vascular bed of the catheter by combining the patient's basic information and derives the impedance change index of the downstream vascular bed of the catheter. Input estimation module: It is used to input the nonlinear waveform complexity index, spectral energy distribution feature vector, pressure-flow velocity coupling coefficient and downstream vascular bed impedance change index as observation sequence into Kalman filter, perform dynamic baseline drift modeling through sliding time window, estimate and generate patient-specific pressure baseline model, and set adaptive catheter occlusion risk benchmark threshold based on the 3σ confidence interval of the prediction error of patient-specific pressure baseline model. The generation module is used to input the nonlinear waveform complexity index, spectral energy distribution feature vector, pressure-flow velocity coupling coefficient, downstream vascular bed impedance change index, and catheter occlusion risk benchmark threshold into a preset Bayesian network. Based on the causal discovery algorithm, it dynamically adjusts the network node connection relationship, constructs a patient-specific risk factor graph structure, and performs causal inference calculation processing in conjunction with the patient's individualized pressure baseline model to output the catheter occlusion probability index. Through recurrent neural network, it performs time-series prediction analysis on the feature sequence to generate a future risk evolution trend index. Finally, it performs hierarchical mapping processing on the catheter occlusion probability index and the future risk evolution trend index to generate a three-level early warning signal including mild occlusion risk, moderate occlusion risk, and severe occlusion risk.

10. The indwelling catheter blockage early warning system based on pressure waveform analysis according to claim 9, characterized in that, The acquisition module includes: The generation unit is set up to establish a multi-sensor spatiotemporal registration architecture, set the sampling frequency parameters of pressure, flow rate and ECG signals, use the IEEE 1588 precise time protocol to achieve nanosecond-level timestamp alignment, and generate original heterogeneous data packets with a unified time base. The original heterogeneous data packets include pressure sampling value sequences, flow rate sampling value sequences and ECG R-wave marker sequences aligned according to the acquisition time. A model unit is established to perform dynamic zero-point compensation of the catheter-blood coupling interface on the original heterogeneous data package. An online calibration algorithm is used to estimate the changes in blood viscosity coefficient and catheter elastic modulus in real time. A pressure baseline drift compensation model is established, and baseline drift is eliminated using this model. Drift-compensated pressure data is then generated. The calculation formula for the pressure baseline drift compensation model is as follows: In the formula, The calculated baseline drift amount, The blood viscosity coefficient was calibrated experimentally. The blood viscosity coefficient is estimated in real time using an online calibration algorithm. This is the preset effective length of the catheter. The elastic deformation coefficient of the catheter was determined experimentally. The elastic modulus of the conduit material is estimated in real time using an online calibration algorithm. For the radial deformation of the duct; Execution unit: Based on the drift-compensated pressure data, the velocity sampling value sequence in the original heterogeneous data package, and the ECG R-wave label sequence, it performs vascular hydrostatic pressure geometric correction and hemodynamic dimension normalization to generate a raw hemodynamic digital twin dataset. The raw hemodynamic digital twin dataset is stored in a circular buffer queue structure and an alignment index is established according to the timestamp.

Images