Methods and systems for evaluating postoperative outcomes of intracranial aneurysm clipping

Abstract

This invention discloses a method and system for evaluating postoperative outcomes after intracranial aneurysm clipping surgery, belonging to the field of medical diagnostics and monitoring technology. It includes acquiring and analyzing real-time multi-source physiological data from postoperative patients, obtaining a dynamic pressure-perfusion correlation map, updating a pre-defined individualized digital model, obtaining the updated individualized digital model results, and performing short-term risk simulations. When the generated short-term risk prediction parameters exceed a pre-defined risk threshold, multivariate simulations of virtual intervention plans are performed to generate simulation results for multiple virtual intervention plans. Based on the dynamic pressure-perfusion correlation map, short-term risk prediction parameters, and simulation results of multiple virtual intervention plans, integrated decision support information is generated. This invention employs the technical means of fusing and analyzing real-time multi-source physiological data to update the individualized digital model and perform risk prediction and virtual intervention simulations, enabling proactive prediction and intervention of postoperative risks and improving the level of individualized management.

Description

Technical Field

[0001] This invention relates to the field of medical diagnostics and monitoring technology, and in particular to a method and system for evaluating the postoperative effects of intracranial aneurysm clipping surgery. Background Technology

[0002] Intracranial aneurysms are a high-risk cerebrovascular disease, and clipping is one of the main treatment methods. Postoperative conditions are complex and variable, often accompanied by serious complications such as cerebral edema, cerebral vasospasm, and delayed ischemia. Therefore, close and precise monitoring of the physiological status of postoperative patients is crucial for improving prognosis. This type of monitoring typically involves continuous monitoring of key physiological indicators such as intracranial pressure and cerebral blood flow, falling under the application scope of medical diagnostic and monitoring equipment.

[0003] In related technologies, Chinese invention patent CN109907732A discloses a method and system for assessing the risk of intracranial aneurysm rupture, comprising: establishing a three-dimensional model including the carrier artery and the aneurysm on the carrier artery based on intracranial imaging data; determining the target morphological parameters of the virtual carrier artery and the virtual aneurysm based on the three-dimensional model; determining the target hemodynamic parameters of the virtual carrier artery and the virtual aneurysm based on the three-dimensional model; and calculating the target morphological parameters, target hemodynamic parameters, and target clinical parameters based on a pre-trained machine learning model to obtain the assessment result of the virtual aneurysm, which is used to assess the risk of aneurysm rupture.

[0004] Regarding the aforementioned technologies, the inventors believe that their assessment models are static and one-off, relying on preoperative images for risk prediction, which fails to meet the needs of postoperative dynamic monitoring. They depend solely on morphological parameters derived from images, resulting in a single data source and failing to reflect the patient's real-time physiological functional state. Furthermore, the pre-trained models used are fixed and lack the individualized ability to dynamically adjust to real-time patient feedback, ultimately providing only risk conclusions without quantitative guidance for intervention decisions. Summary of the Invention

[0005] To address the aforementioned issues, this invention provides a method and system for evaluating the postoperative effects of intracranial aneurysm clipping surgery. It employs a technique of fusing and analyzing real-time multi-source physiological data to update an individualized digital model, and then using this model to perform risk prediction and virtual intervention simulation. This technique generates predictive, individualized integrated decision support information, enabling proactive management and precise intervention of postoperative risks.

[0006] The above objectives can be achieved through the following approach: A method and system for evaluating the postoperative efficacy of intracranial aneurysm clipping surgery includes acquiring real-time multi-source physiological data of postoperative patients, including intracranial pressure waveform data and blood flow monitoring data; fusing the multi-source physiological data and performing pressure-perfusion coupling analysis to obtain a dynamic pressure-perfusion correlation map; updating a preset individualized digital model based on the dynamic pressure-perfusion correlation map to obtain an individualized digital model update result; performing short-term risk extrapolation based on the individualized digital model update result to generate short-term risk prediction parameters; when the short-term risk prediction parameters exceed a preset risk threshold, performing multivariate simulation extrapolation of virtual intervention schemes based on the individualized digital model update result to generate multiple virtual intervention scheme simulation results; and generating integrated decision support information based on the dynamic pressure-perfusion correlation map, the short-term risk prediction parameters, and the multiple virtual intervention scheme simulation results.

[0007] Optionally, acquiring real-time multi-source physiological data of postoperative patients includes: acquiring intracranial pressure waveform data via an intracranial pressure sensor, acquiring blood flow velocity data of major cerebral arteries via a transcranial Doppler ultrasound monitoring device, and acquiring blood oxygen saturation data of local brain tissue via a near-infrared spectral brain oxygen monitoring device; simultaneously acquiring the intracranial pressure waveform data, the blood flow velocity data, and the blood oxygen saturation data, and unifying the blood flow velocity data and the blood oxygen saturation data as blood flow monitoring data, which together with the intracranial pressure waveform data constitute multi-source physiological data.

[0008] Optionally, obtaining the dynamic pressure perfusion correlation map includes: extracting waveform feature parameters from the intracranial pressure waveform data and extracting blood flow feature parameters from the blood flow monitoring data; performing spatiotemporal alignment and correlation analysis on the waveform feature parameters and the blood flow feature parameters; using the changes in the waveform feature parameters as a macroscopic pressure state index; locating and analyzing the spatial or temporal response of the synchronized blood flow feature parameters; and generating the dynamic pressure perfusion correlation map. Optionally, obtaining the personalized digital model update result includes: using the pressure-perfusion coupling relationship and change trend reflected in the dynamic pressure-perfusion association map as a constraint condition input into the personalized digital model; adjusting the vascular reactivity parameters and collateral circulation compensation parameters in the personalized digital model based on the constraint condition to obtain the personalized digital model update result.

[0009] Optionally, generating short-term risk prediction parameters includes: running the individualized digital model to perform forward simulation calculations to generate a brain perfusion state evolution path; and quantifying and extracting the risk probability and expected occurrence time of underperfusion or overperfusion from the brain perfusion state evolution path to generate short-term risk prediction parameters.

[0010] Optionally, generating simulation results of multiple virtual intervention schemes includes: obtaining a preset number of virtual intervention schemes; loading physiological parameter adjustment instructions corresponding to each virtual intervention scheme into the individualized digital model update results; rerunning the individualized digital model loaded with different physiological parameter adjustment instructions to calculate the simulated brain perfusion state after the implementation of each virtual intervention scheme; comparing the differences between the various simulated brain perfusion states and the evolution path of the brain perfusion state to generate simulation results of multiple virtual intervention schemes.

[0011] Optionally, the generation of integrated decision support information includes: superimposing and cross-verifying the spatiotemporal coupling pattern of pressure and perfusion reflected in the dynamic pressure perfusion correlation map, the temporal characteristics of risk evolution revealed by the short-term risk prediction parameters, and the perfusion response trajectories under different intervention paths in the simulation results of the multiple virtual intervention schemes to obtain a superimposed multidimensional situation; based on the superimposed multidimensional situation, identifying key compensation windows and intervention sensitive nodes, and generating integrated decision support information that integrates risk warning, intervention timing recommendation, and expected effect map.

[0012] Optionally, loading the physiological parameter adjustment instructions corresponding to each of the virtual intervention schemes includes: selecting a virtual intervention scheme type, wherein the type includes adjusting the target value of mean arterial pressure, adjusting the infusion rate of vasoactive drugs, or adjusting the intravenous infusion rate; converting the selected virtual intervention scheme type into physiological parameter adjustment instructions for modifying the input boundary conditions or internal parameters of the individualized digital model; and loading the physiological parameter adjustment instructions into the update results of the individualized digital model to simulate the physiological effects of the virtual intervention scheme type.

[0013] Optionally, the method further includes: acquiring the actual intervention plan implemented based on the integrated decision support information and the corresponding actual physiological parameter adjustment amount; monitoring the new real-time multi-source physiological data generated by the patient after implementing the actual intervention plan, and generating a new dynamic pressure perfusion correlation map; calculating the difference between the actual intervention effect reflected by the new dynamic pressure perfusion correlation map and the expected effect of the corresponding plan in the simulation results of the multiple virtual intervention plan, and obtaining the effect deviation; and correcting the internal parameters of the individualized digital model based on the effect deviation.

[0014] Based on the same inventive concept, this invention also provides a postoperative efficacy evaluation system for intracranial aneurysm clipping surgery. The system includes: a data acquisition module for acquiring real-time multi-source physiological data of the postoperative patient, including intracranial pressure waveform data and blood flow monitoring data; a data fusion and analysis module for fusing the multi-source physiological data and performing pressure-perfusion coupling analysis to obtain a dynamic pressure-perfusion correlation map; a model update module for updating a preset individualized digital model based on the dynamic pressure-perfusion correlation map to obtain an individualized digital model update result; a risk extrapolation module for performing short-term risk extrapolation based on the individualized digital model update result to generate short-term risk prediction parameters; a virtual intervention extrapolation module for performing multivariate simulation extrapolation of virtual intervention schemes based on the individualized digital model update result when the short-term risk prediction parameters exceed a preset risk threshold, generating multiple virtual intervention scheme simulation results; and a decision support generation module for generating integrated decision support information based on the dynamic pressure-perfusion correlation map, the short-term risk prediction parameters, and the multiple virtual intervention scheme simulation results.

[0015] Compared with the prior art, the present invention has the following advantages: 1. This invention, through deep fusion and dynamic coupling analysis of multi-source physiological data, transcends the reliance on single parameter thresholds in traditional monitoring, and can reveal the complex and dynamic pathophysiological relationship between intracranial pressure and cerebral blood flow perfusion, thereby providing clinicians with a deeper and more comprehensive understanding of the patient's condition.

[0016] 2. Based on a real-time data-driven individualized digital model, this invention enables quantitative prediction and prospective extrapolation of patients' short-term future risks, transforming postoperative monitoring from a passive event response mode to an active risk warning mode. This can buy valuable treatment windows for clinical intervention and help prevent or reduce the occurrence of secondary brain injury.

[0017] 3. This invention provides doctors with a scientific and quantitative basis for decision-making when formulating treatment strategies by simulating and comparing the effects of various clinical intervention programs in a virtual environment. This evidence-based deduction can help select the optimal intervention path for the current individual patient condition, thereby achieving more precise and individualized treatment management.

[0018] 4. This invention constructs a closed-loop optimization system encompassing data acquisition, model updates, risk prediction, virtual intervention, and feedback correction. By comparing the actual intervention effects with model predictions, it can continuously self-correct and learn, constantly improving the simulation accuracy of physiological responses to specific patients, thus enhancing its assessment and prediction capabilities over time.

[0019] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures pointed out in the description, claims and drawings. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] Figure 1 This is a flowchart illustrating the postoperative effect evaluation method for intracranial aneurysm clipping surgery according to an embodiment of the present invention.

[0022] Figure 2 This is a dynamic pressure perfusion correlation map according to an embodiment of the present invention.

[0023] Figure 3 This is a schematic diagram of short-term risk prediction according to an embodiment of the present invention.

[0024] Figure 4 This is a schematic diagram of the postoperative effect evaluation system for intracranial aneurysm clipping surgery according to an embodiment of the present invention. Detailed Implementation

[0025] 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 embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] Reference Figure 1 One embodiment of the present invention proposes a method for evaluating the postoperative effect of intracranial aneurysm clipping surgery. It employs a technique of fusing and analyzing real-time multi-source physiological data to update an individualized digital model, and then using this model to perform risk prediction and virtual intervention simulation. This method can generate predictive individualized integrated decision support information, enabling proactive management and precise intervention of postoperative risks.

[0027] The method described in this embodiment specifically includes: Real-time multi-source physiological data of postoperative patients are acquired, including intracranial pressure waveform data and blood flow monitoring data. Optionally, acquiring real-time multi-source physiological data of postoperative patients includes: Intracranial pressure waveform data is obtained through an intracranial pressure sensor, blood flow velocity data of major cerebral arteries is obtained through a transcranial Doppler ultrasound monitoring device, and blood oxygen saturation data of local brain tissue is obtained through a near-infrared spectral brain oxygen monitoring device. Simultaneously collect the intracranial pressure waveform data, the blood flow velocity data, and the blood oxygen saturation data, and combine the blood flow velocity data and the blood oxygen saturation data as blood flow monitoring data, which together with the intracranial pressure waveform data constitute multi-source physiological data.

[0028] Specifically, the first step is to standardize the deployment of monitoring equipment and establish data channels. In terms of engineering, a fiber optic or strain gauge intracranial pressure sensor is invasively implanted into the patient's brain parenchyma or ventricle to continuously acquire raw intracranial pressure waveform data at a sampling frequency of at least 100 Hz, ensuring the capture of fine waveform components such as P1, P2, and P3 within the cardiac cycle. Simultaneously, a transcranial Doppler ultrasound probe is fixed at the patient's temporal window to continuously monitor key blood supply vessels such as the middle cerebral artery (MCA), acquiring blood flow velocity envelope data at a frequency of approximately 100 to 200 Hz. Furthermore, a near-infrared spectral brain oxygenation monitoring sensor patch is placed on the patient's forehead to non-invasively monitor local brain tissue oxygen saturation in the frontal cortex; the sampling frequency of this data stream is typically lower, approximately 1 to 2 Hz. To achieve data synchronization, all monitoring devices are connected to a centralized data acquisition and synchronization control unit. This unit incorporates a high-precision master clock. Upon initiation of the acquisition command, it timestamps each frame of data from the intracranial pressure sensor, transcranial Doppler ultrasound monitoring device, and near-infrared spectroscopy cerebral oxygenation monitoring device via hardware triggering or a network time protocol. The timestamp accuracy must be at the millisecond level to ensure precise matching and alignment between high-frequency intracranial pressure waveform data and blood flow velocity data and low-frequency blood oxygen saturation data within the analysis window. Subsequently, the timestamp-aligned blood flow velocity data stream and blood oxygen saturation data stream are uniformly encapsulated in a data structure to form a composite data object, which is defined as blood flow monitoring data. This unified blood flow monitoring data, together with the synchronously acquired intracranial pressure waveform data, constitutes the multi-source physiological data required by this method and is output as a continuous data stream to subsequent steps.

[0029] For example, a post-craniotomy patient undergoing brain hemorrhage surgery is connected to this system in the intensive care unit. First, a neurosurgeon or other professional implants a fiber optic intracranial pressure (ICP) sensor within the patient's ventricle via minimally invasive surgery. This sensor continuously acquires intracranial pressure waveforms at a frequency of 200 Hz. The P1, P2, and P3 components of this waveform reflect brain compliance and vascular resistance. Simultaneously, a transcranial Doppler ultrasound (TCD) probe is fixed to the patient's left temporal window, monitoring the blood flow velocity of the left middle cerebral artery at a frequency of 150 Hz. Furthermore, near-infrared spectroscopy (NIRS) sensors are attached to the patient's left and right foreheads, monitoring local cerebral oxygen saturation (rSO2) at a frequency of 1 Hz. To ensure data synchronization, all devices are connected to a centralized medical data acquisition and synchronization control unit. This unit uses a high-precision master clock and hardware triggering mechanism to assign a uniform millisecond-level timestamp to each frame of data from the ICP, TCD, and NIRS, achieving time alignment of multi-source data. For example, at T=50.000 seconds, ICP waveforms, MCA blood flow velocity, and bilateral rSO2 data can be acquired simultaneously. Subsequently, the time-aligned MCA blood flow velocity data and rSO2 data are encapsulated into a unified "blood flow monitoring data" object. This object, together with the synchronously acquired ICP waveform data, constitutes a multi-source physiological data stream, which is output to subsequent processing modules. By combining invasive and non-invasive methods, multi-dimensional indicators such as intracranial pressure, cerebral blood flow velocity, and cerebral oxygenation status are acquired simultaneously, resolving the data heterogeneity problem between devices with different sampling frequencies and providing a reliable basis for real-time and accurate assessment of the patient's cerebral hemodynamic status.

[0030] By fusing the multi-source physiological data and performing pressure-perfusion coupling analysis, a dynamic pressure-perfusion correlation map is obtained. Optionally, obtaining the dynamic pressure perfusion correlation map includes: Waveform feature parameters are extracted from the intracranial pressure waveform data, and blood flow feature parameters are extracted from the blood flow monitoring data; The waveform feature parameters and the blood flow feature parameters are spatiotemporally aligned and correlated. The changes in the waveform feature parameters are used as a macroscopic pressure state index. The spatial or temporal responses of the synchronized blood flow feature parameters are located and analyzed to generate a dynamic pressure perfusion correlation map.

[0031] Specifically, a parallel feature extraction process is first initiated to process the synchronously input multi-source physiological data. For intracranial pressure waveform data, a set of waveform feature parameters is extracted using time-domain and frequency-domain analysis algorithms within a sliding time window of 10 to 60 seconds. In the time domain, the average intracranial pressure, pulse amplitude (the height ratio of P2 to P1 waves, P2 / P1), and waveform rise time within this window are calculated. In the frequency domain, slow wave activity amplitudes in the range of 0.005 to 0.05 Hz are identified using Fast Fourier Transform. These results constitute the waveform feature parameter set. Simultaneously, for transcranial Doppler ultrasound blood flow velocity data and near-infrared spectral brain oxygen saturation data, which serve as blood flow monitoring data, blood flow feature parameters are extracted within the same time window. This includes obtaining the average blood flow velocity and pulsatility index of the middle cerebral artery, as well as the mean and rate of change of local brain tissue oxygen saturation in the frontal cortex. For calculating the pulsatility index... ,have: ; in, Peak systolic velocity is the highest value of intravascular blood flow velocity during a complete cardiac cycle. End-diastolic velocity is the lowest value of blood flow velocity within a blood vessel during a cardiac cycle. The mean blood flow velocity refers to the time-weighted average of blood flow velocities within a complete cardiac cycle. Subsequently, spatiotemporal alignment and correlation analysis are performed. Spatiotemporal alignment here refers to pairing waveform feature parameters extracted within the same timestamp window with blood flow feature parameters. The core of correlation analysis is to use specific waveform feature parameters, such as the P2 / P1 ratio, as a macroscopic pressure state index. In engineering, changes in this index directly reflect changes in brain compliance; a sustained increase usually indicates a deterioration in intracranial buffering capacity. Using changes in this index as the driving event, cross-correlation or moving window correlation analysis algorithms are employed to calculate its dynamic correlation with synchronized blood flow feature parameters. For example, the Pearson correlation coefficient between specific waveform feature parameters and blood flow feature parameters is calculated using the following formula. : ; in, This represents a time series of waveform characteristic parameters acquired and calculated within a time window T, such as a continuous P2 / P1 ratio sequence. This represents a time series of blood flow characteristic parameters acquired synchronously within the same time window T, such as the average blood flow velocity sequence of the middle cerebral artery. Let be the covariance of these two time series. and These are the standard deviations of sequence X and sequence Y, respectively. The calculated correlation coefficient R is used as a key quantitative indicator. These correlation coefficient R values, which evolve over time and cover different pairs of physiological parameters, are organized and visualized into a matrix or multidimensional map, i.e., a dynamic pressure perfusion correlation map. For example... Figure 2 As shown, the graph is presented in the form of a heatmap. The horizontal axis represents different blood flow characteristic parameters, and the vertical axis represents different time windows. The color intensity represents the Pearson correlation coefficient, or R value, between the P2 / P1 ratio and various blood flow characteristic parameters. The closer the R value is to 1, the stronger the positive correlation; the closer it is to 0, the no correlation; and the closer it is to -1, the stronger the negative correlation.

[0032] For example, for synchronously input intracranial pressure (ICP) waveform data, a sliding time window of 30 seconds is used for analysis. First, characteristic parameters are extracted from the ICP waveform: the average intracranial pressure within the window (e.g., 15 mmHg), the height ratio of P2 to P1 waves (e.g., P2 / P1 = 1.25), and the waveform rise time (e.g., 150 ms). The P2 / P1 ratio is a key indicator; an increase in this ratio usually indicates decreased brain compliance. Simultaneously, frequency domain analysis is performed to extract slow wave amplitudes (e.g., 0.8 mmHg) in the range of 0.005–0.05 Hz, reflecting cerebral vascular autoregulation. Simultaneously, feature extraction is performed on the blood flow monitoring data: the average flow velocity within the window (e.g., 60 cm / s) and pulsatility index are calculated from the middle cerebral artery (MCA) blood flow velocity data. The pulsatility index is calculated to quantify cerebral vascular pulsation patterns. Assuming that within one cardiac cycle, the measured... = 90 cm / s, = 40 cm / s, obtained by time-weighted averaging of blood flow velocity within this cycle. = 65 cm / s. Then calculate the pulsatility index. The results are rounded to two decimal places. This reflects cerebrovascular resistance or compliance. Simultaneously, the mean value of local brain tissue oxygen saturation (rSO2) in the frontal cortex (e.g., 68%) and its rate of change (e.g., a decrease of 0.5% per minute) are calculated. These constitute a set of blood flow characteristic parameters. Spatiotemporal alignment and correlation analysis are then performed. ICP features within the same time window are paired with blood flow characteristic parameters using timestamps, and the P2 / P1 ratio is used as an index for macroscopic pressure status. Through sliding window cross-correlation analysis, the Pearson correlation coefficient between the P2 / P1 sequence and various blood flow parameter sequences, such as the mean flow velocity of the MCA and rSO2, is dynamically calculated. For example, if the calculated covariance is -0.5, and the standard deviations of the two sequences are 0.1 and 5 cm / s, respectively, then... The results indicate a completely negative correlation between increased P2 / P1 and decreased mean MCA flow velocity. A dynamic pressure-perfusion correlation map was created by organizing the correlation coefficients of different parameter pairs over time. This map, with time as the axis, visually displays the response pattern of cerebral perfusion under different pressure states. For example, when P2 / P1 is consistently higher than 1.3, if both MCA flow velocity and rSO2 show a strong negative correlation and R is close to -1, the map will highlight this in red, indicating pressure-dependent perfusion impairment, i.e., an increased risk of cerebral ischemia. Through the above analysis, a real-time mapping of the dynamic coupling relationship between intracranial pressure and cerebral perfusion was achieved. The generated correlation map provides clinicians with an intuitive visualization tool, helping to identify cerebral hemodynamic abnormalities early and providing crucial evidence for intervention decisions.

[0033] Based on the dynamic pressure perfusion correlation map, the preset individualized digital model is updated to obtain the individualized digital model update result. Optionally, obtaining the personalized digital model update result includes: The pressure-infusion coupling relationship and changing trend reflected in the dynamic pressure-infusion association map are used as constraints input into the individualized digital model. Based on the constraints, the vascular reactivity parameters and collateral circulation compensation parameters in the individualized digital model are adjusted to obtain the updated results of the individualized digital model.

[0034] Specifically, the key information extracted from the dynamic pressure-perfusion correlation map is first formatted as constraints for the model. In engineering terms, this means resolving the quantified pressure-perfusion coupling relationship in the map, such as the correlation coefficient R and its trend between changes in specific intracranial pressure waveform characteristic parameters and the blood flow velocity response of the middle cerebral artery, into a set of objective functions or boundary conditions. This individualized digital model is a brain circulation network model constructed based on the lumped parameter method. It simplifies the cerebral vascular system as a circuit network composed of resistive, capacitive, and inductive elements, where vascular reactivity parameters and collateral circulation compensation parameters are key adjustable variables. The vascular reactivity parameter is a dimensionless gain coefficient that controls the strength of the autoregulation mechanism, determining the ability of the simulated blood vessel to respond to pressure changes to maintain perfusion stability. The collateral circulation compensation parameter represents the conductance of bypass vessels such as leptomeningeal anastomoses, determining the compensatory blood flow that the bypass can provide when the main blood flow is obstructed. The update process initiates an iterative optimization algorithm, such as Bayesian inference or nonlinear least squares, to adjust these parameters. A set of simulated pressure perturbations is input into an individualized digital model with current parameters to calculate the simulated perfusion response output by the model, thereby generating a simulated pressure-perfusion correlation. Then, the difference between the simulation results and actual observed data is quantified using the following objective function. The overall error to be minimized in the calculation is... ,have: ; in, Weighting factors assigned to coupling relationships of different clinical importance; The i-th pressure-infusion coupling coefficient simulated by the model; This represents the i-th coupling coefficient value extracted from actual observation data; The value represents the number of key coupling relationships selected from the dynamic pressure perfusion correlation map; the summation symbol Σ indicates the traversal of multiple key coupling relationships selected from the dynamic pressure perfusion correlation map. The optimization algorithm repeatedly adjusts the parameter vector to be optimized, including vascular reactivity parameters and collateral circulation compensation parameters, and recalculates E until E converges to below a preset threshold, typically less than 10 to the power of -4. The parameter set obtained at this point is considered the appropriate solution for the current state. Finally, this set of parameters is solidified into the individualized digital model, forming an updated individualized digital model that can accurately reproduce the patient's current cerebral hemodynamic characteristics, and is then output.

[0035] For example, based on the dynamic pressure-perfusion correlation map obtained above, the pressure-perfusion coupling relationship and its changing trend reflected therein are formatted as constraints for updating the individualized digital model. For instance, if the map shows that the correlation coefficient R between the patient's intracranial pressure P2 / P1 ratio and the mean blood flow velocity in the middle cerebral artery has remained between -0.8 and -1.0 over the past 30 minutes, it indicates that increased intracranial pressure has a significant inhibitory effect on cerebral blood flow, suggesting that the cerebral vascular autoregulation function may be impaired. This strong negative correlation and its trend will be extracted as the target constraint for model optimization. The individualized digital model is a lumped parameter-based cerebral circulation network model, abstracting the vascular system as a circuit network composed of resistance, capacitance, and inductance. The core adjustable parameters of the model include: a vascular reactivity parameter—a dimensionless gain coefficient that controls how blood vessels resist pressure changes to maintain cerebral perfusion, reflecting the strength of autoregulation; and a collateral circulation compensation parameter—representing the conductance value of bypass vessels, determining the compensatory blood flow capacity when the main vessel is blocked. To adjust these parameters, a Bayesian inference optimization algorithm is initiated. The initial model runs with default or preset parameters. During optimization, simulated pressure perturbations, such as transient blood pressure changes, are input into the current model. The perfusion response output by the model is calculated, and the simulated pressure-perfusion correlation, such as the correlation coefficient between simulated P2 / P1 and blood flow velocity, is generated. To quantify the difference between simulation and actual observation, the overall error is calculated and minimized. For example, if the current focus is on the negative correlation between P2 / P1 and blood flow velocity, a higher weight, such as 2.0, can be assigned to it. Assuming only this single key relationship is optimized (N=1), the observed correlation coefficient is -0.9, and the model simulates a value of -0.7 in a certain iteration, then the overall error for that iteration is... The optimization algorithm continuously adjusts parameters and repeatedly calculates the error E until it converges to a preset threshold, such as less than 0.0001. Once convergence is achieved, the resulting parameter set is the optimal solution that accurately reflects the patient's current hemodynamic state. This parameter set is embedded in the model, forming an individualized digital model update result. By transforming real-time coupling relationships into model constraints and performing parameter calibration based on the optimization algorithm, this step achieves accurate updates from a general model to a highly personalized model. This significantly improves the model's predictive accuracy and simulation fidelity, laying a reliable foundation for subsequent risk extrapolation and virtual intervention simulations, thereby enhancing the scientific rigor and effectiveness of clinical decision-making.

[0036] Based on the updated results of the individualized digital model, short-term risk simulation is performed to generate short-term risk prediction parameters; Optionally, the parameters for generating short-term risk prediction include: The individualized digital model is run to perform forward simulation calculations to generate the evolution path of brain perfusion state; From the brain perfusion state evolution path, the risk probability and expected occurrence time of underperfusion or overperfusion are quantitatively extracted to generate short-term risk prediction parameters.

[0037] Specifically, the process begins with a forward simulation based on the updated results of the individualized digital model. Macroscopic physiological parameters such as mean arterial pressure and heart rate at the current moment are used as the initial boundary conditions of the model. It is assumed that these conditions will remain stable or change according to a preset small trend over a future simulation period, such as 6 hours. A numerical solver, such as the fourth-order Runge-Kutta method, is used to integrate the differential equations representing the cerebral hemodynamic network, with a time step set to 0.1 to 1 second to ensure computational stability and accuracy. The calculation results in a series of simulated physiological parameters that evolve over time, covering different vascular network nodes throughout the brain, including simulated cerebral blood flow, cerebrovascular resistance, and capillary pressure in each region. These time-series data collectively constitute the evolution path of cerebral perfusion status. Subsequently, the risk quantification and extraction stage begins. To calculate the risk probability, a Monte Carlo simulation method is used. Based on the above boundary conditions, a random noise perturbation within the physiological range is superimposed, and the simulation is performed N times, typically N being between 500 and 1000 times, with independent simulation operations. For each simulation run, key cerebral perfusion state variables are monitored to see if they reach preset risk thresholds. For example, regional cerebral blood flow below 20 ml / 100 g / min is defined as underperfusion, or above 75 ml / 100 g / min is defined as hyperperfusion. The probability of specific risk events is calculated. ,have: ; in, To monitor the frequency of this risk event; This represents the total number of simulations. For simulations where a risk event occurs, the simulation time point when the risk threshold is first reached is recorded, and the average or median of these time points is calculated as the expected time of occurrence of the risk. For example... Figure 3 As shown in the figure, the curve illustrates the evolution of the probability of cerebral hypoperfusion in patients over the next 6 hours. The curve shows that the risk probability gradually increases over time, reaching and exceeding the preset risk threshold of 60% after 3.2 hours. The risk threshold and the expected time point of high risk are marked in the figure. Finally, a set of quantitative indicators, including the calculated probability of hypoperfusion risk, the expected time of occurrence, and the probability of hyperperfusion risk, the expected time of occurrence, etc., are packaged to generate short-term risk prediction parameters, which are then output for subsequent decision-making.

[0038] For example, forward simulation calculations are initiated based on the updated individualized digital model. First, the current patient's real-time macroscopic physiological parameters, such as mean arterial pressure (MAP) of 85 mmHg and heart rate (HR) of 75 bpm, are used as the initial boundary conditions for the model. It is assumed that these conditions remain stable over a 6-hour simulation period. A high-precision fourth-order Runge-Kutta numerical solver is used to integrate the differential equations representing the cerebral hemodynamic network, with a time step of 0.5 seconds to ensure computational stability and accuracy. The calculation results generate a series of simulated physiological parameters that evolve over time, covering different vascular network nodes throughout the brain, including simulated cerebral blood flow in various regions, such as blood flow in the left MCA supply area, cerebral vascular resistance, and capillary pressure. These time-series data collectively constitute a detailed evolutionary path of cerebral perfusion status. Subsequently, the risk quantification and extraction stage begins. To calculate the probability of underperfusion or overperfusion, a Monte Carlo simulation method is employed. Based on the aforementioned boundary conditions, a slight random noise perturbation within the physiological range is superimposed in each simulation step, resulting in 1000 independent simulations. For each simulation run, key cerebral perfusion state variables, such as blood flow in the left MCA supply area, are monitored to see if they reach a preset risk threshold. For example, if the simulated blood flow is below 20 ml / 100 g / min, it is defined as underperfusion; if it is above 75 ml / 100 g / min, it is defined as overperfusion. The probability of a specific risk event, such as underperfusion, is calculated. Assuming that in 1000 Monte Carlo simulations, 150 simulations predict that the patient will experience a left MCA supply area blood flow below 20 ml / 100 g / min (i.e., an underperfusion event) within the next 6 hours, the probability of underperfusion risk is calculated. The threshold is 15%. For each simulation of a risk event, the time point at which the risk threshold is first reached is recorded. For example, in 150 simulations predicting hypoperfusion, the time point at which hypoperfusion first occurs is recorded, such as 2.5 hours in the first simulation, 3.1 hours in the second simulation, and so on. Then, the median of these time points, such as 3.2 hours, is calculated as the expected time of occurrence of the risk. Finally, a set of quantitative indicators, such as the calculated hypoperfusion risk probability (e.g., 15%) and expected time of occurrence (e.g., 3.2 hours), and the hyperperfusion risk probability (e.g., 5%) and expected time of occurrence (e.g., 4.8 hours), are packaged to generate short-term risk prediction parameters, which are then highlighted and output to the clinician's workstation to indicate the patient's potential cerebral perfusion risk. By combining the updated individualized digital model and Monte Carlo simulation, this step can prospectively predict the probability and expected time of occurrence of cerebral hypoperfusion or hyperperfusion in the patient within a future period, such as 6 hours. This enables clinicians to shift from a reactive approach to a proactive one, allowing them to identify high-risk patients in advance, thus gaining valuable time for preventative interventions or adjustments to treatment plans and reducing the risk of postoperative brain injury.

[0039] When the short-term risk prediction parameters exceed the preset risk threshold, based on the updated results of the individualized digital model, a multivariate simulation of the virtual intervention plan is performed to generate simulation results of multiple virtual intervention plans. Optionally, the simulation results of generating multiple virtual intervention schemes include: Obtain a preset number of virtual intervention schemes, and load the physiological parameter adjustment instructions corresponding to each virtual intervention scheme into the individualized digital model update results; The individualized digital model, loaded with different physiological parameter adjustment instructions, was re-run, and the simulated brain perfusion state after the implementation of each virtual intervention scheme was calculated. By comparing the differences between various simulated brain perfusion states and the evolution paths of those brain perfusion states, simulation results of multiple virtual intervention schemes are generated.

[0040] Specifically, the process begins by retrieving a pre-defined number of clinically common virtual intervention protocols, typically 3 to 5, from a pre-defined library. Each virtual intervention protocol is then converted into specific physiological parameter adjustment instructions and loaded into the updated results of the individualized digital model. For example, a protocol to "increase mean arterial pressure to 90 mmHg" is converted into an instruction to modify the target value of the arterial pressure driving function in the model's input boundary conditions to 90 mmHg. Subsequently, for the individualized digital model loaded with each physiological parameter adjustment instruction, the forward simulation is restarted. This calculation process is consistent with the process of generating the brain perfusion state evolution path, running for the same duration and outputting a new simulated brain perfusion state time series for each intervention protocol. Finally, a differential analysis is performed to quantify the potential effect of each intervention protocol. This is achieved by calculating the difference between the simulated brain perfusion state after intervention and the original brain perfusion state evolution path, generating simulation results for multiple virtual intervention protocols. This difference can be quantified using an effect deviation index. The effect deviation produced by the i-th virtual intervention protocol... ,have: ; in, The key variable values ​​of simulated brain perfusion status at simulated time point t after implementing the i-th intervention scheme, such as regional cerebral blood flow. The values ​​of the corresponding variables in the brain perfusion state evolution path at the same time point t without any intervention are given. T represents the total simulation duration. This integral calculates the overall deviation between the intervention path and the baseline path throughout the entire simulation period. The final simulation result of the multiple virtual intervention schemes is a set of effect deviations from all measured schemes, as well as a visual comparison atlas of the simulated brain perfusion state time series under each scheme, jointly presenting the advantages and disadvantages of different intervention paths.

[0041] For example, if short-term risk prediction parameters indicate a 70% probability of hypoperfusion in the patient within the next 6 hours, exceeding a preset threshold, a simulation of multiple virtual intervention protocols will be performed based on an updated individualized digital model. Multiple preset clinical protocols, such as "increasing mean arterial pressure to 90 mmHg" or "using vasoactive drugs," will be loaded and translated into precise adjustment instructions for the model's boundary conditions or internal parameters. Subsequently, the model independently runs each protocol, generating a corresponding simulated cerebral perfusion state evolution path, which is compared with the baseline path of "no intervention." To quantify the effect, an effect deviation index is calculated, which measures the overall difference between each intervention path and the risk baseline path. For example, if cerebral blood flow would drop to 15 ml / 100 g / min without intervention, while an intervention protocol can maintain it at 30 ml / 100 g / min, then the effect deviation of that protocol is high, representing a significant intervention effect. Finally, a simulation result of multiple virtual intervention protocols, including the effect deviation of each protocol and a visual comparison atlas, is generated. The results visually present the expected effects and advantages and disadvantages of different interventions, providing clinicians with a tool for "sand table simulation" before actual operation, helping them to choose the personalized treatment strategy with the least risk and the best effect, thereby avoiding blind trial and error and improving the accuracy and safety of treatment.

[0042] Based on the dynamic pressure perfusion correlation map, the short-term risk prediction parameters, and the simulation results of the multiple virtual intervention schemes, integrated decision support information is generated.

[0043] Optionally, the generation of integrated decision support information includes: The dynamic pressure perfusion correlation map reflects the spatiotemporal coupling mode of pressure and perfusion, the short-term risk prediction parameters reveal the temporal characteristics of risk evolution, and the perfusion response trajectory under different intervention paths in the simulation results of the multiple virtual intervention schemes are superimposed and cross-verified to obtain the superimposed multidimensional situation. Based on the superimposed multidimensional situation, key compensation windows and sensitive intervention nodes are identified, and integrated decision support information that combines risk warning, intervention timing recommendations, and expected effect maps is generated.

[0044] Specifically, this step first executes a multi-dimensional situation overlay and cross-verification procedure. In a unified graphical user interface, using a shared timeline as a baseline, a dynamic pressure-perfusion correlation map, short-term risk prediction parameters, and simulation results of multiple virtual intervention schemes are visualized and overlaid. In this interface, the dynamic pressure-perfusion correlation map displays the spatiotemporal distribution pattern of the coupling strength between pressure and perfusion parameters in the form of a heatmap, both at the current moment and over a past period. The short-term risk prediction parameters, in the form of risk probability curves, indicate the evolutionary time series characteristics of the risk of insufficient or excessive brain perfusion on the future timeline, clearly marking the key time points when the risk value exceeds the warning threshold. The simulation results of multiple virtual intervention schemes are overlaid on the risk curve as multiple perfusion response trajectory curves of different colors, visually demonstrating the corrective effect of different intervention paths on the future risk trajectory. Subsequently, this overlaid multi-dimensional situation is automatically analyzed to identify critical compensatory windows and intervention-sensitive nodes. The algorithm for identifying the critical compensatory window period searches along the timeline to pinpoint the time interval from the current moment until the short-term risk prediction parameter first reaches the high-risk threshold, where the simulation results of multiple virtual intervention schemes show the most significant intervention effect. The identification of intervention-sensitive nodes is achieved through cross-validation. This involves analyzing the physiological parameter pairs exhibiting the most significant anomalous coupling in the dynamic pressure-perfusion correlation map and matching them with the simulation results of multiple virtual intervention schemes. The physiological parameter regulation instruction corresponding to the virtual intervention scheme that can most effectively improve this anomalous coupling is identified as the intervention-sensitive node. Finally, based on the above analysis results, an integrated report is automatically generated. This report first explicitly raises a risk warning, such as "The probability of insufficient perfusion in the left middle cerebral artery supply area is 85% within the next 6 hours"; then it provides an intervention timing recommendation, i.e., the identified critical compensatory window period, such as "It is recommended to initiate intervention within the next 2 hours"; finally, it includes an expected effect map, which is a labeled multi-dimensional situational visualization interface that clearly compares the expected perfusion response trajectories of the "no intervention" path and each "virtual intervention" path. This comprehensive report, which integrates risk warnings, intervention timing recommendations, and expected effect maps, is the final integrated decision support information and is presented to clinical users.

[0045] For example, at a certain time point, such as 24 hours post-surgery, the generation of integrated decision support information is initiated. In an integrated and visualized decision support interface, multi-dimensional information is dynamically integrated and presented: a dynamic pressure-perfusion correlation map is displayed in the form of a heatmap, revealing in real time abnormal coupling patterns such as the continuously increasing negative correlation between P2 / P1 and blood flow velocity in the right middle cerebral artery region, with an R value reaching -0.95; simultaneously, short-term risk prediction parameters are marked with risk probability curves, clearly showing that the risk of insufficient perfusion in the right blood supply area will increase from 10% to 75% in the next 4 hours, and exceed the warning threshold after 3.5 hours; furthermore, the simulation results of multiple virtual intervention programs are displayed by overlaying trajectory curves of different colors, allowing for a direct comparison of the effects of each program—for example, the "increasing mean arterial pressure to 90 mmHg" program can reduce the risk from 75% to 20% after 4 hours, while the "using vasoactive drugs" program can only reduce it to 45%. Based on the above overlay situation, the system automatically analyzes and identifies key compensatory windows, such as the next 1-2 hours, when the intervention effect is most significant and the intervention sensitive nodes, such as "mean arterial pressure," are the most effective regulatory targets. Ultimately, an integrated report is automatically generated, including clear risk warnings such as "75% risk of right-sided hypoperfusion within the next 3.5 hours," intervention timing recommendations such as "It is recommended to initiate intervention within the next 1-2 hours," and an annotated expected effect graph that visually compares the differences between no intervention and various virtual intervention pathways, clearly indicating the optimal solution such as "increasing mean arterial pressure to 90 mmHg." By intelligently integrating and visualizing multi-source data, this integrated decision support information not only provides clear risk warnings and quantitative predictions, but also assists physicians in quickly and accurately formulating personalized intervention strategies by identifying critical windows and sensitive points, thereby improving treatment precision and safety.

[0046] Optionally, loading the physiological parameter adjustment instructions corresponding to each of the virtual intervention schemes includes: Select a type of virtual intervention protocol, which includes adjusting the target value of mean arterial pressure, adjusting the infusion rate of vasoactive drugs, or adjusting the intravenous infusion rate; The selected virtual intervention scheme type is converted into physiological parameter adjustment instructions for modifying the input boundary conditions or internal parameters of the individualized digital model; The physiological parameter adjustment instructions are loaded into the updated results of the individualized digital model to simulate the physiological effects of the virtual intervention scheme.

[0047] Specifically, this step begins by presenting a pre-defined menu of virtual intervention options to the user interface. The user selects one or more options for simulation, such as adjusting the target mean arterial pressure (MAP), adjusting the infusion rate of vasoactive drugs, or adjusting the intravenous infusion rate. Once the user makes a selection, a conversion module is immediately activated, translating the selection into one or a set of physiological parameter adjustment instructions. For example, selecting to adjust the MAP target value to 95 mmHg will be translated into an instruction to modify the average value of the pressure source function representing the systemic circulatory inlet in the individualized digital model to 95 mmHg—a direct modification of the model's input boundary conditions. Adjusting the infusion rate of vasoactive drugs, such as norepinephrine, to 0.1 μg / kg / min, will invoke a built-in pharmacokinetic and pharmacodynamic sub-model to calculate its impact on the model's internal parameters based on the input drug type and dosage. The calculation results in new values ​​for model internal parameters, such as systemic vascular resistance, after drug effect adjustment. ,have: ; in, This is the original baseline value of this parameter in the personalized digital model update results before loading the drug instruction; This represents the maximum change in this parameter that the drug can cause; it is a drug-specific constant obtained from a pharmacology database. The selected drug infusion rate is the input variable of the virtual intervention protocol; This is the drug dose that produces 50% of the maximum effect, characterizing drug sensitivity, and is also provided by the database; The Hill coefficient describes the steepness of the dose-response curve. Using this formula, qualitative drug adjustment strategies are precisely quantified as specific numerical modifications to model intrinsic parameters, such as vascular resistance or compliance parameters. Finally, these generated physiological parameter adjustment instructions, whether modifying boundary conditions or intrinsic parameters, are loaded and applied to the updated results of the individualized digital model, preparing for subsequent physiological effect simulations.

[0048] For example, a pre-defined menu of virtual intervention types is first presented to the clinical user interface. This menu contains several options, such as "Adjust target mean arterial pressure" and "Adjust vasoactive drug infusion rate." Clinicians can select one or more options for simulation based on the patient's condition. For example, a doctor might select "Adjust vasoactive drug infusion rate" and want to adjust the norepinephrine infusion rate from the current 0.03 μg / kg / min to 0.05 μg / kg / min. Once the user makes and confirms this selection, a conversion module is immediately activated, translating this selection into one or a set of precise physiological parameter adjustment instructions. For norepinephrine infusion rate adjustment, a built-in pharmacokinetic and pharmacodynamic sub-model is invoked. This sub-model calculates the effect of the input drug type (e.g., norepinephrine) and target dose (e.g., 0.05 μg / kg / min) on personalized digital model internal parameters, such as total systemic vascular resistance. The calculation uses the new values ​​of the model's internal parameters, such as total systemic vascular resistance, after drug effect adjustment. = 2000 dynes per second per centimeter 5 , = 1000 dynes·second / cm 5 , It is 0.05 micrograms per kilogram per minute. = 0.04 micrograms / kg / min = 2, then dynes per second per centimeter 5 The results are rounded to two decimal places. Finally, these generated physiological parameter adjustment instructions, whether modifying boundary conditions or internal parameters, are loaded and applied to the updated results of the individualized digital model, preparing for subsequent physiological effect simulations. By introducing biophysical models such as pharmacokinetic / pharmacodynamic sub-models and Hill equations, vague clinical instructions are transformed into quantifiable model parameter adjustments. This transformation mechanism allows physicians to simulate diverse clinical interventions, rather than being limited to simple parameter adjustments.

[0049] Optionally, the method further includes: Obtain the actual intervention plan implemented based on the integrated decision support information and the corresponding actual physiological parameter adjustment amounts; After the actual intervention program was implemented, the patient's new real-time multi-source physiological data were monitored, and a new dynamic pressure perfusion correlation map was generated. The difference between the actual intervention effect reflected by the new dynamic pressure perfusion correlation map and the expected effect of the corresponding scheme in the simulation results of the multiple virtual intervention scheme is calculated to obtain the effect deviation; Based on the aforementioned performance deviation, the internal parameters of the individualized digital model are corrected using feedback.

[0050] Specifically, this step first uses a recording module to capture the actual intervention plan implemented based on integrated decision support information and its corresponding adjustments to actual physiological parameters; for example, if a clinician adopts a suggestion to "increase the target mean arterial pressure to 95 mmHg." Subsequently, new real-time multi-source physiological data generated by the patient after implementing the actual intervention plan are continuously monitored and acquired. Following the same engineering process described above, this new data is processed to generate a new dynamic pressure-perfusion correlation map reflecting the true physiological state after the intervention. Next, the core bias calculation is performed. Key pressure-perfusion coupling indicators are extracted from this new dynamic pressure-perfusion correlation map as a quantitative representation of the actual intervention effect. Simultaneously, the expected effect of the virtual intervention plan corresponding to the actual intervention plan is retrieved from the previously generated simulation results of multiple virtual intervention plans. The bias between the two effects is then calculated. ,have: ; in, In the new dynamic pressure-perfusion correlation map, the core parameter reflecting the coupling relationship between actual pressure and perfusion after intervention is, such as the correlation coefficient between the new P2 / P1 ratio after intervention and the mean blood flow velocity of the middle cerebral artery. In the simulation results of multiple virtual intervention schemes, this refers to the simulated value of the same coupling parameter predicted by the corresponding virtual intervention scheme. Finally, the calculated effect bias is used as the objective function input to a feedback correction algorithm, such as gradient descent or extended Kalman filter. This algorithm iterates repeatedly by fine-tuning deep internal parameters in the individualized digital model, such as the brain tissue compliance time constant or the gain coefficient of the autoregulation mechanism—parameters that cannot be directly clinically intervened but affect the overall behavior of the model—until the effect bias is minimized to a preset convergence threshold, such as less than 0.05. After this process, the updated model parameters are saved, thus achieving closed-loop feedback correction of the individualized digital model, enabling it to have higher fidelity in the next round of risk simulation.

[0051] For example, after a clinician adopts and implements an intervention based on previously generated integrated decision support information, such as a warning of insufficient left cerebral perfusion, suggesting an increase in mean arterial pressure (MAP) to 90 mmHg within the next hour, a closed-loop feedback correction mechanism is initiated. First, the implemented intervention and its corresponding physiological parameter adjustments are captured. For example, it is recorded that the physician increased the patient's MAP from 80 mmHg to the target value of 90 mmHg at 10:30 AM by adjusting the infusion rate of vasopressors. Subsequently, new real-time multi-source physiological data generated by the patient after implementing this intervention are continuously monitored and acquired. Following the same engineering process described above, this new data is processed to generate a "new dynamic pressure-perfusion correlation map" reflecting the true physiological state after the intervention. This new map shows the changes in the coupling relationship between the ICP waveform P2 / P1 ratio and parameters such as the mean blood flow velocity (CBFV) of the MCA after the MAP increase. For example, the new map shows that after the MAP increase, the negative correlation between P2 / P1 and MCA CBFV weakens from -0.9 to -0.5, while MCA CBFV increases. Next, we perform the core bias calculation. Key pressure-perfusion coupling indicators are extracted from this new dynamic pressure-perfusion correlation map as a quantitative representation of the actual intervention effect. For example, we focus on the new correlation coefficient between P2 / P1 and the mean MCA blood flow velocity after MAP increase. Simultaneously, from the previously generated multiple virtual intervention simulation results, we retrieve the predicted expected effect of the virtual protocol corresponding to the actual intervention protocol of "increasing mean arterial pressure to 90 mmHg". For example, the correlation coefficient between P2 / P1 and the mean MCA blood flow velocity predicted by this virtual protocol is [value missing]. If, after the actual intervention, MAP increases to 90 mmHg, by analyzing the new dynamic pressure-perfusion correlation map, we obtain an actual correlation coefficient of -0.55 between P2 / P1 and the mean MCA blood flow velocity. However, in the virtual intervention simulation, the same protocol with MAP increased to 90 mmHg, the predicted correlation coefficient is -0.60. Therefore, we calculate the effect bias. Finally, the calculated performance bias is used as the objective function input to a feedback correction algorithm, such as an extended Kalman filter. By fine-tuning the deep internal parameters of the individualized digital model—for example, those parameters that cannot be directly intervened in clinical practice but affect the overall behavior of the model, such as the brain tissue compliance time constant and the upper and lower thresholds of cerebrovascular autoregulation mechanisms—the process is iterated repeatedly until the performance bias is minimized to a preset convergence threshold, such as less than 0.02. After this process, the updated model parameters are saved. By constructing a key closed-loop feedback mechanism, the individualized digital model can continuously learn and evolve through the validation of actual clinical interventions and error correction.

[0052] Based on the same inventive concept, such as Figure 4As shown, the present invention also provides a postoperative outcome evaluation system for intracranial aneurysm clipping surgery, the system comprising: The data acquisition module is used to acquire real-time multi-source physiological data of postoperative patients, including intracranial pressure waveform data and blood flow monitoring data. The data fusion and analysis module is used to fuse the multi-source physiological data and perform pressure and perfusion coupling analysis to obtain a dynamic pressure-perfusion correlation map. The model update module is used to update the preset individualized digital model based on the dynamic pressure perfusion correlation map to obtain the individualized digital model update result. The risk simulation module is used to perform short-term risk simulation based on the updated results of the individualized digital model and generate short-term risk prediction parameters. The virtual intervention simulation module is used to perform multivariate simulation of virtual intervention schemes based on the updated results of the individualized digital model when the short-term risk prediction parameters exceed the preset risk threshold, and generate simulation results of multiple virtual intervention schemes. The decision support generation module is used to generate integrated decision support information based on the dynamic pressure perfusion correlation map, the short-term risk prediction parameters, and the simulation results of the multiple virtual intervention schemes.

[0053] It should be noted that the electrical connections between the various units described above do not necessarily represent direct or indirect connections. Any indirect connection method can be applied to the embodiments of the present invention as long as it achieves the purpose of the present invention. The above descriptions are merely exemplary embodiments of the present invention and should not be construed as limiting the scope of the present invention.

[0054] All equivalent changes and modifications made in accordance with the teachings of this invention are still within the scope of this invention. Those skilled in the art will readily conceive of other embodiments of this invention upon considering the specification and the disclosure of practical truth. This application is intended to cover any variations, uses, or adaptations of this invention that follow the general principles of this invention and include common knowledge or conventional techniques in the art not described herein.

Claims

1. A method for evaluating the postoperative effect of intracranial aneurysm clipping surgery, characterized in that, The method includes: Real-time multi-source physiological data of postoperative patients are acquired, including intracranial pressure waveform data and blood flow monitoring data. By fusing the multi-source physiological data and performing pressure-perfusion coupling analysis, a dynamic pressure-perfusion correlation map is obtained. Based on the dynamic pressure perfusion correlation map, the preset individualized digital model is updated to obtain the individualized digital model update result. Based on the updated results of the individualized digital model, short-term risk simulation is performed to generate short-term risk prediction parameters; When the short-term risk prediction parameters exceed the preset risk threshold, based on the updated results of the individualized digital model, a multivariate simulation of the virtual intervention plan is performed to generate simulation results of multiple virtual intervention plans. Based on the dynamic pressure perfusion correlation map, the short-term risk prediction parameters, and the simulation results of the multiple virtual intervention schemes, integrated decision support information is generated.

2. The method for evaluating the postoperative effect of intracranial aneurysm clipping surgery according to claim 1, characterized in that, The acquisition of real-time multi-source physiological data of postoperative patients includes: Intracranial pressure waveform data is obtained through an intracranial pressure sensor, blood flow velocity data of major cerebral arteries is obtained through a transcranial Doppler ultrasound monitoring device, and blood oxygen saturation data of local brain tissue is obtained through a near-infrared spectral brain oxygen monitoring device. Simultaneously collect the intracranial pressure waveform data, the blood flow velocity data, and the blood oxygen saturation data, and combine the blood flow velocity data and the blood oxygen saturation data as blood flow monitoring data, which together with the intracranial pressure waveform data constitute multi-source physiological data.

3. The method for evaluating the postoperative effect of intracranial aneurysm clipping surgery according to claim 1, characterized in that, The obtained dynamic pressure perfusion correlation map includes: Waveform feature parameters are extracted from the intracranial pressure waveform data, and blood flow feature parameters are extracted from the blood flow monitoring data; The waveform feature parameters and the blood flow feature parameters are spatiotemporally aligned and correlated. The changes in the waveform feature parameters are used as a macroscopic pressure state index. The spatial or temporal responses of the synchronized blood flow feature parameters are located and analyzed to generate a dynamic pressure perfusion correlation map.

4. The method for evaluating the postoperative effect of intracranial aneurysm clipping surgery according to claim 1, characterized in that, The obtained personalized digital model update results include: The pressure-infusion coupling relationship and changing trend reflected in the dynamic pressure-infusion association map are used as constraints input into the individualized digital model. Based on the constraints, the vascular reactivity parameters and collateral circulation compensation parameters in the individualized digital model are adjusted to obtain the updated results of the individualized digital model.

5. The method for evaluating the postoperative effect of intracranial aneurysm clipping surgery according to claim 1, characterized in that, The parameters for generating short-term risk prediction include: The individualized digital model is run to perform forward simulation calculations to generate the evolution path of brain perfusion state; From the brain perfusion state evolution path, the risk probability and expected occurrence time of underperfusion or overperfusion are quantitatively extracted to generate short-term risk prediction parameters.

6. The method for evaluating the postoperative effect of intracranial aneurysm clipping surgery according to claim 5, characterized in that, The simulation results of generating multiple virtual intervention schemes include: Obtain a preset number of virtual intervention schemes, and load the physiological parameter adjustment instructions corresponding to each virtual intervention scheme into the individualized digital model update results; The individualized digital model, loaded with different physiological parameter adjustment instructions, was re-run, and the simulated brain perfusion state after the implementation of each virtual intervention scheme was calculated. By comparing the differences between various simulated brain perfusion states and the evolution paths of those brain perfusion states, simulation results of multiple virtual intervention schemes are generated.

7. The method for evaluating the postoperative effect of intracranial aneurysm clipping surgery according to claim 1, characterized in that, The generated integrated decision support information includes: The dynamic pressure perfusion correlation map reflects the spatiotemporal coupling mode of pressure and perfusion, the short-term risk prediction parameters reveal the temporal characteristics of risk evolution, and the perfusion response trajectory under different intervention paths in the simulation results of the multiple virtual intervention schemes are superimposed and cross-verified to obtain the superimposed multidimensional situation. Based on the superimposed multidimensional situation, key compensation windows and sensitive intervention nodes are identified, and integrated decision support information that combines risk warning, intervention timing recommendations, and expected effect maps is generated.

8. The method for evaluating the postoperative effect of intracranial aneurysm clipping surgery according to claim 6, characterized in that, The loading of physiological parameter adjustment instructions corresponding to each of the virtual intervention schemes includes: Select a type of virtual intervention protocol, which includes adjusting the target value of mean arterial pressure, adjusting the infusion rate of vasoactive drugs, or adjusting the intravenous infusion rate; The selected virtual intervention scheme type is converted into physiological parameter adjustment instructions for modifying the input boundary conditions or internal parameters of the individualized digital model; The physiological parameter adjustment instructions are loaded into the updated results of the individualized digital model to simulate the physiological effects of the virtual intervention scheme.

9. The method for evaluating the postoperative effect of intracranial aneurysm clipping surgery according to claim 1, characterized in that, The method further includes: Obtain the actual intervention plan implemented based on the integrated decision support information and the corresponding actual physiological parameter adjustment amounts; After the actual intervention program was implemented, the patient's new real-time multi-source physiological data were monitored, and a new dynamic pressure perfusion correlation map was generated. The difference between the actual intervention effect reflected by the new dynamic pressure perfusion correlation map and the expected effect of the corresponding scheme in the simulation results of the multiple virtual intervention scheme is calculated to obtain the effect deviation; Based on the aforementioned performance deviation, the internal parameters of the individualized digital model are corrected using feedback.

10. A postoperative outcome evaluation system for intracranial aneurysm clipping surgery, applied to the postoperative outcome evaluation method for intracranial aneurysm clipping surgery as described in any one of claims 1-9, characterized in that, The system includes: The data acquisition module is used to acquire real-time multi-source physiological data of postoperative patients, including intracranial pressure waveform data and blood flow monitoring data. The data fusion and analysis module is used to fuse the multi-source physiological data and perform pressure and perfusion coupling analysis to obtain a dynamic pressure-perfusion correlation map. The model update module is used to update the preset individualized digital model based on the dynamic pressure perfusion correlation map to obtain the individualized digital model update result. The risk simulation module is used to perform short-term risk simulation based on the updated results of the individualized digital model and generate short-term risk prediction parameters. The virtual intervention simulation module is used to perform multivariate simulation of virtual intervention schemes based on the updated results of the individualized digital model when the short-term risk prediction parameters exceed the preset risk threshold, and generate simulation results of multiple virtual intervention schemes. The decision support generation module is used to generate integrated decision support information based on the dynamic pressure perfusion correlation map, the short-term risk prediction parameters, and the simulation results of the multiple virtual intervention schemes.

Images