A cerebral ischemic injury regulation analysis system

Abstract

This invention relates to the field of healthcare informatics technology and discloses a brain ischemia injury regulation and analysis system, comprising: using the pressure fluctuation component generated by endogenous physiological activity in the arterial pressure signal as a detection signal, and analyzing the response characteristics of the intracranial pressure signal to this detection signal; and when the detection signal intensity is insufficient, determining the respiratory activity signal as a substitute detection signal for continuous analysis. This invention establishes a data processing method that differs from traditional static threshold monitoring. By analyzing the dynamic response relationship between core parameters, it constructs an assessment dimension that directly reflects the system's intrinsic regulatory capacity. It can reveal the depletion of functional reserves before the absolute values ​​of conventional physiological parameters such as intracranial pressure become abnormal, thereby gaining time for clinical intervention.

Description

Technical Field

[0001] This invention relates to a brain ischemia injury regulation and analysis system, belonging to the field of medical and health informatics technology. Background Technology

[0002] Currently, a common technical strategy in neurocritical care is to use sensors to continuously monitor patients' core physiological parameters, such as intracranial pressure and mean arterial pressure, and compare the monitored values ​​with preset safety thresholds. The application of healthcare informatics in this field is mainly reflected in integrating these independent data streams into a unified monitoring platform for clinical observation and decision-making.

[0003] However, when this monitoring strategy is applied to assess the true functional reserves of a patient's brain under apparent stable physiological parameters, the limitations in information acquisition can affect subsequent clinical judgment. During intensive care, a patient's various physiological parameters may all be within the normal range, but this apparent stable data may correspond to two completely different functional states: the brain's autonomic regulatory function is intact or it is on the verge of failure. Existing monitoring strategies can only reflect the current state of the system, but lack the technical dimension to assess how much disturbance the system can still withstand. As a result, the initiation of clinical intervention measures often requires waiting for an abnormality in the absolute value of a certain parameter before it can be triggered.

[0004] To address this problem, a direct approach is to improve the precision of signal processing or introduce more complex statistical models. However, these methods typically focus on smoothing the signal to filter out fluctuations and obtain a more stable mean. This approach, in turn, discards information about the system's dynamic response capabilities contained in the inherent variability of the signal. Specifically, existing data processing methods suffer from the following limitations: 1. Their analytical models are usually based on linear system assumptions, failing to fully consider the brain's key nonlinear functional reserves; 2. Their processing typically filters out the inherent variability of physiological signals as noise, thus losing early information contained in the signal's complexity or entropy that can predict changes in the system's functional state; 3. Their analytical logic focuses on the independent values ​​of each signal, failing to generate assessment information about the overall robustness of the system from the dynamic coupling relationships of multiple signals. The combined effect of these limitations means that, in clinical situations requiring precise assessment of functional reserves to support intervention, existing data analysis methods, due to the inherent limitations of their processing logic, cannot provide effective information support. Therefore, the technical problem to be solved by this invention is how to use existing conventional multimodal continuous monitoring data to create a data analysis system that can go beyond traditional static threshold comparisons and construct an assessment index that can be used to quantify brain functional reserves and system vulnerability by processing and utilizing the dynamic response characteristics and intrinsic variability information ignored in the signal. Summary of the Invention

[0005] This invention provides a brain ischemia injury regulation and analysis system, the main purpose of which is to solve the problem that the existing data processing methods can only perform static threshold comparisons and cannot obtain information on the dynamic functional reserves of the brain and the vulnerability assessment of the system from routine monitoring data.

[0006] To achieve the above objectives, the present invention provides a brain ischemia injury regulation and analysis system, comprising:

[0007] The signal acquisition module is configured to simultaneously acquire mean arterial pressure time-series signals and intracranial pressure time-series signals;

[0008] The detection signal extraction module is configured to extract pressure fluctuations in the frequency range of 0.05Hz to 0.15Hz generated by vascular autoregulation by performing bandpass filtering on the mean arterial pressure time series signal, and define the pressure fluctuations as endogenous detection signals;

[0009] The detection source determination module is configured to monitor the signal energy of the endogenous detection signal. If the signal energy is not continuously lower than an activation threshold, the endogenous detection signal is determined as the current detection signal. If the signal energy is continuously lower than the activation threshold, the respiratory cycle signal extracted from the synchronously acquired respiratory activity signal is determined as the current detection signal.

[0010] The response feature calculation module is configured to calculate the phase delay and amplitude gain of the intracranial pressure response based on the determined current detection signal and the intracranial pressure time series signal corresponding to the current detection signal; wherein, when the endogenous detection signal is the current detection signal, a short time window is set with the time points of each peak and each trough of the endogenous detection signal as the center, and the calculation is performed based on the endogenous detection signal and the corresponding intracranial pressure time series signal within the short time window;

[0011] The assessment index generation module is configured to generate assessment indicators characterizing the brain's functional reserve state based on the determined current detection signal and the intracranial pressure time series signal corresponding to the current detection signal.

[0012] Preferably, the activation threshold is predetermined based on the statistical energy distribution of the endogenous probe signal in the historical baseline mean arterial pressure time series signal.

[0013] Preferably, the signal acquisition module is further configured to: acquire synchronized EEG time-series signals; calculate the sample entropy of the EEG time-series signals within a short time window or within a period corresponding to the respiratory cycle signal; and generate evaluation indicators by combining the synchronization measures of phase delay, amplitude gain, and sample entropy through a fixed linear combination model.

[0014] Preferred evaluation indicators Generated according to the following rules: in, As evaluation indicators, The normalized phase delay, The normalized amplitude gain. As a normalized measure of synchronicity, , and These are linear weighting coefficients that are preset based on the predictive contribution of each input feature in historical sample data and whose sum is 1.

[0015] Preferably, the system is further configured to, after determining the respiratory cycle signal as the current detection signal and before step F, include: acquiring an airway pressure waveform generated by the breathing device and synchronized with the respiratory cycle signal, and using the airway pressure waveform as a template signal; and performing matched filtering on the intracranial pressure time series signal using the template signal to obtain an enhanced intracranial pressure response signal, and calculating the phase delay and amplitude gain using the enhanced intracranial pressure response signal.

[0016] Preferably, the system further includes: analyzing the EEG time-series signal to detect whether there are preset artifact features caused by non-physiological physical interference; and suspending the operation of the evaluation index generation module during the time period in which the artifact features are detected.

[0017] Preferably, the system is further configured to: classify the detected artifact features into one of the categories of impact artifacts and rhythmic artifacts; if the artifact category is impact artifacts, select an impact response template subtraction model; if the artifact category is rhythmic artifacts, select an adaptive notch filter model; process the mean arterial pressure time series signal and intracranial pressure time series signal using the selected model to obtain a corrected signal; and perform a response feature calculation module based on the corrected signal.

[0018] Preferably, the system further includes: calculating a current mean arterial pressure level by performing low-pass filtering on the mean arterial pressure time series signal; based on the mean arterial pressure level, selecting a matching evaluation model parameter set from multiple preset evaluation model parameter sets corresponding to three arterial pressure level intervals (hypotension, normal blood pressure, and hypertension) for use by the evaluation index generation module; the multiple preset evaluation model parameter sets are independently trained and calibrated using historical datasets corresponding to the corresponding arterial pressure level intervals.

[0019] Preferably, the system further includes: obtaining a cerebral perfusion pressure index by calculating the difference between the mean arterial pressure time series signal and the intracranial pressure time series signal; and generating a final regulatory analysis result by combining the assessment index and the cerebral perfusion pressure index according to a preset arbitration rule, wherein the arbitration rule stipulates that when the assessment index indicates that the brain functional reserve is good, but the cerebral perfusion pressure index is lower than the survival threshold of 60 mmHg, the final regulatory analysis result is judged as a critical condition.

[0020] Compared with the prior art, the beneficial effects of the present invention are:

[0021] 1. This invention utilizes the pressure fluctuation component in arterial pressure signals caused by endogenous physiological activities as a probe signal continuously applied to brain tissue, and couples and analyzes the response characteristics of intracranial pressure signals to this probe signal, establishing a data processing method that differs from traditional static threshold monitoring. This method no longer judges the value of a single physiological parameter in isolation, but constructs an assessment dimension that directly reflects the system's intrinsic regulatory capacity by analyzing the dynamic response relationship between two core parameters. Changes in this dimension can reveal the state of functional reserve depletion before the absolute values ​​of conventional physiological parameters such as intracranial pressure or arterial pressure show obvious abnormalities.

[0022] 2. This invention also provides a switching mechanism for detection signal sources. When the intensity of intrinsic arterial pressure fluctuations, which serve as active detection signals, is insufficient, the periodic components generated by respiratory activity on intracranial pressure signals are automatically identified as alternative detection signals. By reusing the data fluctuations associated with a common life support measure (mechanical ventilation) as a backup information detection source and processing them within a unified response analysis framework, the assessment of brain functional reserves can continue even under specific conditions where active detection sources are weakened or disappear due to clinical intervention, thus avoiding the loss of critical monitoring information.

[0023] 3. This invention further integrates the analysis of electroencephalogram (EEG) signals into the assessment process, but its purpose is not to directly assess neurological function. Instead, it serves as a basis for identifying physical interference. When artifact features caused by non-physiological physical interference appear in the EEG signals, the intracranial pressure and arterial pressure response analysis results corresponding to the artifact in time are marked or discarded. This processing method utilizes the different response characteristics of signals from different sources to the same physical event, establishing an independent verification channel for the core hemodynamic response analysis, distinguishing between the real physiological regulatory response and the pseudo-response caused by external interference, and improving the reliability of the final assessment indicators.

[0024] 4. Through the combination of the above mechanisms, this invention forms a complete information processing chain: it extracts endogenous detection and response signals from conventional monitoring data to construct the core logic for evaluating functional reserves, and addresses the special working condition of missing active detection signals through seamless switching of backup detection sources. Then, it uses third-party signals to identify and purify data quality in real time, eliminating the confusion caused by physical interference. The organic combination of these three links enables this system to stably output a dynamic evaluation index with internal logical consistency regarding the regulatory capacity of the brain tissue system from the continuously collected raw multimodal data stream mixed with noise and interference. Attached Figure Description

[0025] Figure 1 This is a flowchart of the brain ischemia injury regulation and analysis system of the present invention;

[0026] Figure 2 This is a diagram showing the weight coefficient configuration of the evaluation model for different blood pressure zones in this invention;

[0027] Figure 3 This is a diagram of the data processing and information interaction architecture of the control and analysis system of the present invention. Detailed Implementation

[0028] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be described in detail below. However, those skilled in the art will understand that the specification and embodiments of the present invention are intended to explain the present invention and not to limit the scope of protection of the present invention.

[0029] This invention provides a brain ischemia injury regulation and analysis system. In terms of information processing architecture, it mainly includes a signal acquisition module, a probe signal extraction module, a probe source determination module, a response feature calculation module, and an evaluation index generation module. The system is configured as a continuous data processing link, receiving multimodal physiological time-series signals as input, and outputting an evaluation index characterizing the brain's functional reserve state through a series of signal processing and calculation procedures. In the information platform of a neurocritical care unit, the static threshold comparison method of continuously monitoring data streams has limitations in revealing the depletion of functional reserves in an apparent steady state. Therefore, the data processing method of this invention is configured to utilize the inherent variability of physiological signals, using the pressure fluctuation component of arterial pressure signals generated by physiological activity as a probe signal applied to brain tissue, and coupling and analyzing the response characteristics of intracranial pressure signals to this probe signal. An evaluation method reflecting the system's regulatory capacity is established. Specifically, the system's signal acquisition module is configured to synchronously acquire mean arterial pressure (MAP) time-series signals and intracranial pressure (ICP) time-series signals from the monitoring device's data interface. To separate the detectable dynamic information from the MAP signal, the detection signal extraction module uses a bandpass filtering algorithm to process the MAP time-series signal. This algorithm is designed to only detect pressure fluctuations in the frequency range of 0.05Hz to 0.15Hz generated by vascular autoregulation. The filtered pressure fluctuation time series is defined internally as an endogenous detection signal. For example, a bandpass filter can be set as a fourth-order Butterworth filter with a passband frequency between 0.05Hz and 0.15Hz. When a raw MAP time-series signal is input into this filter, its output is a pressure fluctuation signal containing only this specific frequency component.

[0030] After obtaining the endogenous probe signal, to quantify the brain tissue's response pattern to the probe signal, the response feature calculation module executes the following procedures: First, it identifies the time points of each peak and trough in the endogenous probe signal, and sets a short time window of fixed width, for example, 5 seconds, centered on these time points; Second, within this short time window, the system simultaneously analyzes the morphological changes of the intracranial pressure time series signal and calculates two response feature parameters, namely the phase delay and amplitude gain of the intracranial pressure response; wherein, the phase delay is calculated as the intracranial pressure signal... The time difference between the occurrence of a peak or trough in the intracranial pressure signal and the corresponding peak or trough in the endogenous probe signal is calculated, while the amplitude gain is calculated as the ratio of the peak-trough difference in the intracranial pressure signal within that time window to the peak-trough difference in the corresponding endogenous probe signal. By processing continuous endogenous probe signal cycles, the system can generate two continuously varying time-series parameters: phase delay and amplitude gain. In clinical interventions, such as when using vasoactive drugs, the fluctuation intensity in the mean arterial pressure time-series signal may weaken, leading to a decrease in the signal energy of the endogenous probe signal. Insufficient for reliable response analysis; to address this situation and ensure the continuous operation of the analysis system, this invention sets up a detector source determination module, which implements a detector source switching mechanism based on signal energy monitoring. This module continuously calculates the signal energy of the endogenous detector signal within a sliding time window and compares it with a preset activation threshold. The determination of the activation threshold follows a standardized calibration procedure, namely, in the initial operation phase of the system, it is calculated based on the statistical energy distribution of the endogenous detector signal in the historical baseline mean arterial pressure time series signal, and a specific quantile of this distribution, such as the 5th percentile, is set as the activation threshold. During system operation, if the signal energy is continuously lower than the activation threshold, for example, for 3 consecutive minutes, the detector source determination module determines that the main detector source has failed and automatically determines the respiratory cycle signal extracted from the synchronously acquired respiratory activity signal as the current detector signal. Subsequent response feature calculations are based on this alternative detector signal, thereby utilizing the data fluctuations associated with measures such as mechanical ventilation as a backup information detector source and avoiding the loss of monitoring information.

[0031] To further enhance the robustness of analysis under conditions of low signal-to-noise ratio of surrogate probe signals, such as when poor patient brain compliance leads to a weak response of respiratory waves in intracranial pressure signals, the system is also configured to perform a template-matching-based signal enhancement process. Under this procedure, after the respiratory cycle signal is identified as the current probe signal, the system further acquires an airway pressure waveform synchronized with the respiratory cycle signal from the data interface of the respiratory device, and uses this airway pressure waveform as the template signal. Subsequently, matched filtering is performed on the original intracranial pressure time series signal using the template signal. This process amplifies the morphologically correlated components of the intracranial pressure signal with the template signal while suppressing irrelevant noise, thereby generating a... An enhanced intracranial pressure response signal is generated, and subsequent phase delay and amplitude gain calculations are based on this enhanced signal. Furthermore, to integrate multi-dimensional information to generate the final evaluation index, the evaluation index generation module is configured to execute a multivariate linear combination model. In this model, in addition to the aforementioned phase delay and amplitude gain, a third information dimension is introduced: the sample entropy calculated by analyzing synchronously acquired EEG time-series signals, and a measure of the synchronicity between this sample entropy and hemodynamic response characteristics. Here, sample entropy, as a technical indicator characterizing the complexity of EEG signals, is calculated following a standard algorithm. Finally, the CFRI evaluation index, characterizing the brain's functional reserve state, is generated according to the following rules: In the formula, The phase delay after normalization. The normalized amplitude gain. As a normalized measure of synchronicity, , and These are linear weighting coefficients that are preset based on the predictive contribution of each input feature in historical sample data and sum to 1. These weighting coefficients are obtained by independent training and calibration using historical datasets of the corresponding arterial pressure level intervals in an offline state.

[0032] In a monitored environment, non-physiological physical disturbances, such as patient movement or nursing procedures, can simultaneously cause fluctuations in mean arterial pressure and intracranial pressure signals, generating spurious response events. To distinguish between physiological regulatory responses and spurious responses caused by external disturbances, this system further integrates artifact feature analysis of EEG time-series signals. The system is configured to monitor in parallel whether preset artifact features caused by non-physiological physical disturbances exist in the EEG time-series signals, such as short-duration high-amplitude and full-band power synchronous bursts caused by muscle activity or electrode displacement. In one implementation, during the time period when artifact features are detected, the system suspends the generation step of evaluation indicators. Alternatively, the system may mark the evaluation indicators calculated within that time period as invalid. In another implementation, instead of halting the calculation, the system first classifies the detected artifact features into one of two categories: impact artifacts or rhythmic artifacts. If the artifact category is impact artifacts, such as those caused by coughing, the system selects an impact response template subtraction model. If the artifact category is rhythmic artifacts, such as those caused by chills, the system selects an adaptive notch filter model. Subsequently, the system uses the selected model to process the mean arterial pressure time series signal and the intracranial pressure time series signal to obtain a corrected signal, and performs the evaluation indicator generation step based on the corrected signal.

[0033] Considering the nonlinear nature of the brain's autoregulatory function, its response pattern varies with background arterial pressure levels. To improve the physiological accuracy of the assessment model, this system also incorporates an adaptive dynamic switching mechanism for assessment models based on pressure zones. Under this mechanism, the system first calculates a current mean arterial pressure level by low-pass filtering the mean arterial pressure time-series signal. Then, based on this mean arterial pressure level, it selects a matching assessment model parameter set from multiple preset parameter sets corresponding to three arterial pressure level zones: hypotension, normal blood pressure, and hypertension. For example, when the mean arterial pressure level is below 60 mmHg, the parameter set of the hypotension model is loaded. Finally, the selected assessment model parameter set is used to generate the assessment index. Finally, to ensure the safety of the system output results and the reliability of clinical interpretation, this system also integrates a decision arbitration gating mechanism. This mechanism performs an independent security check before outputting the final regulatory analysis results. Specifically, the system obtains a cerebral perfusion pressure index by calculating the difference between the mean arterial pressure time-series signal and the intracranial pressure time-series signal in parallel. Based on a preset arbitration rule, and in conjunction with the assessment index... Together with the cerebral perfusion pressure index, a final regulatory analysis result is generated. The arbitration rule stipulates that when the assessment index indicates that the brain functional reserve is good, but the cerebral perfusion pressure index is lower than a preset survival threshold, such as 60 mmHg, the system will determine the final regulatory analysis result as a critical condition. This arbitration mechanism provides a safety boundary for the output of the main protocol, ensuring that the system's analysis conclusions do not violate basic physiological principles.

[0034] Example 1: In a neurocritical care unit, the monitoring information system of a patient with acute cerebral ischemia-reperfusion injury showed that the mean arterial pressure time series signal fluctuated within the range of 70-80 mmHg, and the intracranial pressure time series signal fluctuated within the range of 15-18 mmHg. The simultaneously acquired EEG time series signal did not show any clear epileptic waves, and the absolute values ​​of all parameters were within the set safety thresholds. Based on this, a conventional monitoring system assessed the patient's condition as stable. When the regulation and analysis system of this invention was applied to this scenario, it did not directly use the absolute values ​​of the above parameters for judgment, but instead began to dynamically analyze the multimodal data stream. Response relationship analysis: First, the system's detection signal extraction module performs bandpass filtering on the acquired mean arterial pressure time series signal, separating a pressure fluctuation with a frequency range between 0.05Hz and 0.15Hz, and defining this fluctuation as an endogenous detection signal. Simultaneously, the response feature calculation module uses the peaks and troughs of this endogenous detection signal as a time reference, analyzing the corresponding response in the synchronized intracranial pressure time series signal, calculating the initial phase delay as 1.2 seconds and the amplitude gain as 0.15. The evaluation index generation module further combines the sample entropy calculation results of the EEG signal to output the brain function reserve index. The initial value is 0.8.

[0035] Over time, although the absolute values ​​of the patient's mean arterial pressure and intracranial pressure did not change significantly and remained within safe thresholds, the analysis system of this invention detected changes in intrinsic dynamic response characteristics. Specifically, the phase delay between the intrinsic detection signal and the intracranial pressure response signal gradually increased to 2.5 seconds, and the amplitude gain rose to 0.30, reflecting a decrease in the intracranial pressure's buffering capacity against arterial pressure fluctuations, and a more direct pressure transmission. Simultaneously, the evaluation index generation module analysis showed that the sample entropy of the EEG time-series signal exhibited a decreasing trend, indicating a reduction in the signal's intrinsic complexity. This change occurred synchronously with changes in hemodynamic response characteristics. Based on these changes in multidimensional response characteristics, the brain functional reserve index calculated by the evaluation index generation module... The intracranial pressure continued to decline and fell below the preset alarm threshold of 0.4 after several hours. This early warning information, based on dynamic relationship analysis, was pushed to the clinical monitoring interface by the system. Its occurrence was earlier than the routine alarm event of the absolute value of intracranial pressure exceeding 20 mmHg. The system provides information that is different from static parameter values ​​by outputting assessment indicators that reflect regulatory capacity, thus solving the problem of assessment uncertainty when apparent physiological stability and intrinsic functional failure coexist. The system's early warning output provides a time reference for clinicians to take prospective intervention measures, helping to avoid delays in intervention after waiting for abnormal static physiological parameters to appear. Finally, by changing the way monitoring data is processed and using the existing processor in the monitoring system to perform analysis, early identification of the risk of secondary brain injury is achieved.

[0036] Example 2: To objectively verify the effectiveness of the technical solution of the present invention, this example conducted a retrospective data analysis experiment based on publicly available physiological database records. The aim was to quantitatively compare the performance of the analysis system using the method of the present invention with that using conventional static threshold monitoring in identifying events of declining brain functional reserve. The experimental data came from multimodal physiological signal records in the PhysioNet publicly available database. Data segments containing synchronous and continuous mean arterial pressure time-series signals, intracranial pressure time-series signals, and electroencephalogram (EEG) time-series signals were selected, with a sampling rate of no less than 100Hz during data acquisition. The experiment included an experimental group using the method of the present invention and a control group simulating conventional monitoring. The processor in the experimental group was configured to execute the complete analysis process, including extracting endogenous probe signals, calculating phase delay and amplitude gain, and fusing EEG sample entropy to generate a brain functional reserve index. and set A reading below 0.4 was considered a warning event. The processor in the control group was configured to monitor only the absolute value of the intracranial pressure time-series signal, with a warning event set for intracranial pressure consistently above 20 mmHg. Both processors processed the same selected data segment in parallel, covering the period from a stable physiological state to the onset of increased intracranial pressure. In the initial phase of the experiment, the intracranial pressure fluctuated around 16 mmHg, and the mean arterial pressure fluctuated around 75 mmHg. Neither system issued a warning. The experimental group's system at this point calculated a phase delay of 1.3 seconds, an amplitude gain of 0.14, a sample entropy of 1.8, and a corresponding brain function reserve index. The value was 0.75. As data processing progressed, after a certain time point T0, the intermediate parameters calculated by the experimental group system began to show a trend change. See Table 1 for the specific data evolution.

[0037] Table 1: Examples of the evolution of key parameters at different time points.

[0038]

[0039] According to the data in Table 1, during the time period from T0 to T0+75 min, the absolute values ​​of intracranial pressure and mean arterial pressure did not reach the 20 mmHg threshold. Therefore, the control group system did not issue any warnings during this period. The experimental group system, through analysis of the dynamic response relationship of the signals, monitored a unidirectional increase in phase delay and amplitude gain, as well as a continuous decrease in sample entropy. These changes in intermediate characteristics collectively led to the brain functional reserve index... The index steadily declined, and at T0+45 min, it first fell below 0.4, triggering an early warning event in the experimental group system. Subsequently, at T0+90 min, the intracranial pressure value in the raw data rose to 22.5 mmHg, at which point the control group system triggered an early warning event based on a static threshold for the first time. The experimental results show that for the same physiological data, the early warning event generated by the experimental group using the method of this invention occurred at T0+45 min, while the early warning event of the control group using the conventional static threshold monitoring method occurred at T0+90 min, a difference of 45 minutes. This data confirms that the technical solution of this invention can identify the depletion of functional reserves before the absolute value of conventional monitoring parameters becomes abnormal through quantitative analysis of the dynamic response relationship of multimodal physiological signals.

[0040] Example 3: This example combines Figures 1 to 3 This describes a system for analyzing the regulation of brain ischemia-reperfusion injury, such as... Figure 1 As shown, the system begins by simultaneously acquiring multimodal signals such as mean arterial pressure, intracranial pressure, and respiration. Bandpass filtering is then performed to extract pressure fluctuations with frequencies between 0.05 Hz and 0.15 Hz as detection signals. The primary or backup detection source is determined based on whether the energy of this detection signal remains below a threshold. The primary detection source is the endogenous pressure fluctuation, while the backup detection source is the respiratory cycle signal. After determining the detection signal, the system calculates response characteristics in parallel: phase delay and amplitude gain are calculated based on the current detection signal and intracranial pressure signal; EEG sample entropy is calculated to analyze the complexity of the EEG signal; and brain perfusion pressure is calculated for subsequent safety arbitration. Subsequently, the response characteristics and sample entropy are fused to generate the evaluation index CFRI, which then enters the decision arbitration stage. If the brain perfusion pressure is below the survival threshold, the final result is determined to be a critical condition; otherwise, the evaluation index is output as the final result.

[0041] like Figure 2 As shown, it displays the three linear weighting coefficients used to calculate the evaluation metric CFRI, namely the phase delay weight, in the form of a bar chart. Amplitude gain weight With sample entropy weight Preset values ​​within three different arterial pressure ranges, with a weighting of <65 mmHg in the low blood pressure range, are as follows: =0.55、 =0.30 and =0.15, within the normal blood pressure range of 65-90 mmHg, the weight is adjusted to... =0.30、 =0.40 and =0.30; while in the hypertension range >90 mmHg, the weight is set to 0.30. =0.15、 =0.55 and =0.30.

[0042] like Figure 3 As shown, the raw physiological signal stream originates from the bedside monitoring device. This signal stream is synchronized and forwarded via a signal acquisition gateway and then sent to a data processing server through a secure data transmission protocol. The core engine of the regulation and analysis system runs on this server. The data processing server interacts with the data storage server through a database connection protocol. The latter contains a physiological database, a historical baseline database, an artifact feature database, and an evaluation model database. Finally, the data processing server transmits the analysis results to the visualization monitoring client on the clinician's workstation via a real-time result push protocol.

[0043] Example 4: This example illustrates a standardized offline calibration procedure for the core algorithm model and key judgment thresholds of the technical solution before deployment. This procedure aims to determine parameters in the solution that partially rely on historical data statistical characteristics through a deterministic process. These parameters include those used to generate the brain functional reserve index. linear weighting coefficients , , The calibration procedure includes an activation threshold for activating backup detection signal sources. The initial input is a dataset containing multiple sets of anonymized historical monitoring records. Each set contains at least 24 hours of synchronously acquired mean arterial pressure time-series, intracranial pressure time-series, and electroencephalogram (EEG) time-series signals. Each set also includes a binary clinical outcome label indicating whether the individual experienced a secondary brain injury event within the following 72 hours. The calibration process is deployed and executed on a data processing server. The first step of the procedure is to establish a pre-defined artifact feature library for identifying physical interference. The processing program first iterates through all EEG time-series signals in the dataset, and technicians annotate time segments that clearly correspond to physical interference events based on signal morphology and synchronous nursing records. Subsequently, the program automatically extracts features from these annotated segments, calculating their statistical characteristics in the time and frequency domains, such as signal amplitude exceeding the patient's historical... The duration of the baseline at 6 standard deviations and the ratio of energy in the 10Hz to 40Hz frequency band to energy in the 0.5Hz to 4Hz frequency band were used to determine the statistical distribution of these features. These features were then stored in a queryable database to form a pre-defined artifact feature library. At time T0+60min, the monitoring system detected a persistent interference with narrow-band frequency characteristics in the EEG time-series signal. The artifact feature analysis module within the system compared this interference with the pre-defined artifact feature library and classified it as a rhythmic artifact. Instead of halting the calculation, the system automatically loaded an adaptive notch filter model into the data processing flow for this time period. This model, centered on the identified interference frequency, filtered the synchronized mean arterial pressure time-series signal and intracranial pressure time-series signal. Subsequent phase delay and amplitude gain calculations were based on the corrected signal output by this model, thus eliminating the interference from the physical interference in the final evaluation index calculation without interrupting monitoring.

[0044] The second step of the procedure is to determine the activation threshold used for probe source switching. The processing program uses the dataset purified in the first step to extract the mean arterial pressure time-series signal of the first 60 minutes for each group of records as baseline data. For each baseline data segment, the program performs bandpass filtering to obtain the endogenous probe signal and calculates the average energy of this signal using a 1-minute sliding window. All window energy values ​​from all records are aggregated to form an overall energy distribution database. The activation threshold is ultimately set at the 10th percentile of all energy values ​​in this database. The third step of the procedure is to calibrate the index used to calculate the brain functional reserve index. linear weighting coefficients , and The processing procedure first performs a complete response feature calculation on all purified records in the dataset, generating instantaneous values ​​of phase delay, amplitude gain, and sample entropy for each time point. Then, using the clinical outcome label as the dependent variable and the statistical values ​​of the response feature parameters in the corresponding time series as independent variables, a multivariate logistic regression model is constructed. After model training, the regression coefficients corresponding to the three independent variables (phase delay, amplitude gain, and sample entropy) are extracted and normalized so that their sum equals 1. The three normalized values ​​are then determined as linear weight coefficients. , and By executing the above complete calibration procedure, the key algorithm parameters in the technical solution are determined through a reproducible process based on objective data and statistical models, providing engineering feasibility for the deployment and application of the analysis system across different information platforms.

[0045] Example 5: This example illustrates the collaborative operation procedure of the adaptive evaluation model dynamic switching mechanism and the decision arbitration gating mechanism of the analysis system when facing hemodynamic boundary conditions. In a data segment used for system stress testing, it represents a hypotensive state where the mean arterial pressure drops from 80 mmHg to 55 mmHg within 10 minutes, while the intracranial pressure remains stable at 15 mmHg. To ensure the reliability of the system's analysis under such conditions, the system pre-calibrates a dedicated set of evaluation model parameters for different arterial pressure level intervals. The calibration process is as follows: first, the historical dataset is divided into three subsets based on the mean arterial pressure level: hypotension, normal blood pressure, and hypertension. Then, for each data subset, multivariate logistic regression and coefficient normalization steps are performed independently to generate a set of corresponding linear weight coefficients for each pressure interval.

[0046] At the start of the stress test, the system detected that the mean arterial pressure level was within the normal blood pressure range. Therefore, the parameter set of the assessment model for the normal blood pressure range was loaded to calculate the brain functional reserve index. When the mean arterial pressure in a data segment decreases and remains below the 65 mmHg threshold, the system automatically switches to and loads the parameter set of the specific evaluation model calibrated for the hypotension data subset mentioned above, in order to continue generating data. When the mean arterial pressure drops to 55 mmHg, the calculated cerebral perfusion pressure index has decreased to 40 mmHg, below the survival threshold of 60 mmHg. At this point, even the values ​​calculated by the hypotension model... Even if the value is not lower than the alarm threshold of 0.4, the system's decision arbitration gating mechanism is still triggered, and a control analysis result that determines the state to be critical is finally output, thus ensuring that the system has a layer of safety verification based on basic physiological principles in addition to the calculation of core indicators.

[0047] Example 6: To enable the analysis system to adapt to different monitoring objects, the system is configured to execute an online initialization and baseline adaptive calibration procedure when starting a new monitoring session. When the analysis system first connects to the data interface of a new monitoring object and starts, the system first enters an online initialization phase that lasts for 30 minutes. During this phase, the system continuously acquires the initial data segments of the mean arterial pressure time series signal, intracranial pressure time series signal, and electroencephalogram (EEG) time series signal of the monitoring object.

[0048] The system processor uses this 30-minute data set to execute a pre-defined baseline establishment algorithm. The steps are as follows: First, calculate the standard deviation of the EEG signal amplitude within the data segment, and set six times this standard deviation as the individualized artifact detection amplitude threshold for this monitoring session. Second, the system calculates the stable average phase delay and amplitude gain within this data segment, and uses these two values ​​as the initial baseline reference values ​​for this monitoring session. Only after the initialization phase is completed does the system begin to formally calculate and output continuously changing data. To adapt to changes in individual physiological states during long-term monitoring, the system also implements a background baseline update mechanism. This mechanism uses a 24-hour sliding time window to cache response characteristic parameter data and compares the statistical distribution of the most recent 12 hours' data with the initial baseline reference value every 12 hours. If the change in the average phase delay exceeds the 20% drift threshold, the system automatically uses the data within the current 24-hour sliding window as the new baseline to update the initial baseline reference value.

[0049] To further verify the necessity and technical advantages of the dual-probe switching mechanism proposed in this invention in dealing with common clinical interventions, the following comparative example 1 is set up.

[0050] Comparative Example 1: This comparative example aims to simulate a technical solution that relies solely on a single endogenous detection signal for dynamic analysis. Its analysis method is completely consistent with the experimental group in Example 2 in terms of signal acquisition, bandpass filtering, and calculation rules for response characteristics (phase delay, amplitude gain, sample entropy). The essential difference is that the analysis method used in this comparative example does not include the switching mechanism of this invention, which automatically determines the respiratory cycle signal as the current detection signal when the endogenous detection signal energy is continuously lower than the activation threshold, and interrupts the calculation and output of evaluation indicators when it detects insufficient endogenous detection signal energy. Using the method of this comparative example, a retrospective analysis was performed on the data of another group of patients with similar clinical backgrounds. During the monitoring process, due to hemodynamic instability, at time point T0+30min, the clinician began to continuously administer vasoactive drugs (norepinephrine) via an intravenous pump to maintain the mean arterial pressure within the target range. The processor of this comparative example method and the processor of the control group in Example 2 processed this data in parallel, and the evolution process of its key parameters is recorded in Table 2.

[0051] Table 2: Examples of the evolution of key parameters at different time points in Comparative Example 1.

[0052]

[0053] According to the data in Table 2, after the administration of vasoactive drugs at T0+30 min, the signal energy of intrinsic pressure fluctuations, the primary detection signal, began to decrease significantly. By T0+60 min, this signal energy had decreased to 0.19, consistently below the system's preset activation threshold of 0.25. This caused the comparative analysis system to be forced to interrupt calculations due to the lack of an effective detection signal, and thus could not continue to output the evaluation index CFRI. During subsequent monitoring, although the patient's intracranial pressure eventually rose to 21.5 mmHg at T0+105 min, triggering the conventional static threshold alarm, During the critical 45 minutes from T0+60min to T0+105min, the system in this comparative example was in an information-blinded state and failed to provide any assessment information on the brain's functional reserve status. The experimental results show that in the absence of the dual-source switching mechanism described in this invention, the conventional single-source dynamic analysis method will experience monitoring interruption when faced with common situations where endogenous signals are weakened due to clinical intervention (such as the use of vasoactive drugs). This interruption prevents the system from continuously assessing the risk of subsequent secondary brain injury, thus losing the value of gaining a time window for clinical intervention.

[0054] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0055] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A system for regulating and analyzing cerebral ischemia-reperfusion injury, characterized in that, include: The signal acquisition module is configured to simultaneously acquire mean arterial pressure time-series signals and intracranial pressure time-series signals; The detection signal extraction module is configured to extract pressure fluctuations in the frequency range of 0.05Hz to 0.15Hz generated by vascular autoregulation by performing bandpass filtering on the mean arterial pressure time series signal, and define the pressure fluctuations as endogenous detection signals; The detection source determination module is configured to monitor the signal energy of the endogenous detection signal. If the signal energy is not continuously lower than an activation threshold, the endogenous detection signal is determined as the current detection signal. If the signal energy is continuously lower than the activation threshold, the respiratory cycle signal extracted from the synchronously acquired respiratory activity signal is determined as the current detection signal. The response feature calculation module is configured to calculate the phase delay and amplitude gain of the intracranial pressure response based on the determined current detection signal and the intracranial pressure time series signal corresponding to the current detection signal; wherein, when the endogenous detection signal is the current detection signal, a short time window is set with the time points of each peak and each trough of the endogenous detection signal as the center, and the calculation is performed based on the endogenous detection signal and the corresponding intracranial pressure time series signal within the short time window; The assessment index generation module is configured to generate assessment indicators characterizing the brain's functional reserve state based on the determined current detection signal and the intracranial pressure time series signal corresponding to the current detection signal.

2. The brain ischemia injury regulation and analysis system according to claim 1, characterized in that, The activation threshold is predetermined based on the statistical energy distribution of the endogenous probe signal in the historical baseline mean arterial pressure time series signal.

3. The brain ischemia injury regulation and analysis system according to claim 1, characterized in that, The signal acquisition module is also configured to acquire synchronized EEG time-series signals; calculate the sample entropy of the EEG time-series signals within a short time window or within a period corresponding to the respiratory cycle signal; and generate evaluation indicators by combining phase delay, amplitude gain, and sample entropy synchronization measures through a fixed linear combination model.

4. The brain ischemia injury regulation and analysis system according to claim 2, characterized in that, Evaluation indicators Generated according to the following rules: ,in, As evaluation indicators, The normalized phase delay, The normalized amplitude gain. As a normalized measure of synchronicity, , and These are linear weighting coefficients that are preset based on the predictive contribution of each input feature in historical sample data and whose sum is 1.

5. The brain ischemia injury regulation and analysis system according to claim 1, characterized in that, The system is further configured to, after determining the respiratory cycle signal as the current detection signal, acquire an airway pressure waveform generated by the breathing device that is synchronized with the respiratory cycle signal, and use the airway pressure waveform as a template signal; and perform matched filtering on the intracranial pressure time series signal using the template signal to obtain an enhanced intracranial pressure response signal, and calculate the phase delay and amplitude gain using the enhanced intracranial pressure response signal.

6. The brain ischemia injury regulation and analysis system according to claim 3, characterized in that, The system also includes: analyzing EEG time-series signals to detect whether there are preset artifact features caused by non-physiological physical interference; and suspending the operation of the evaluation index generation module during the time period in which artifact features are detected.

7. The brain ischemia injury regulation and analysis system according to claim 6, characterized in that, The system is further configured to: classify detected artifact features into one of two categories: impact artifacts and rhythmic artifacts; if the artifact category is impact artifacts, select an impact response template subtraction model; if the artifact category is rhythmic artifacts, select an adaptive notch filter model; and process the mean arterial pressure time series signal and intracranial pressure time series signal using the selected model to obtain a corrected signal. The response feature calculation module performs calculations based on the calibrated signal.

8. The brain ischemia injury regulation and analysis system according to claim 1, characterized in that, The system also includes: calculating a current mean arterial pressure level by performing low-pass filtering on the mean arterial pressure time series signal; based on the mean arterial pressure level, selecting a matching evaluation model parameter set from multiple preset evaluation model parameter sets corresponding to three arterial pressure level intervals (hypotension, normal blood pressure, and hypertension) for use by the evaluation index generation module; the multiple preset evaluation model parameter sets are independently trained and calibrated using historical datasets corresponding to the corresponding arterial pressure level intervals.

9. The brain ischemia injury regulation and analysis system according to claim 1, characterized in that, The system also includes: obtaining a cerebral perfusion pressure index by calculating the difference between the mean arterial pressure time series signal and the intracranial pressure time series signal; and generating a final regulatory analysis result by combining the assessment index and the cerebral perfusion pressure index according to a preset arbitration rule. The arbitration rule stipulates that when the assessment index indicates that the brain functional reserve is good, but the cerebral perfusion pressure index is lower than the survival threshold of 60 mmHg, the final regulatory analysis result will be judged as a critical condition.

Images