AI-based ercp sedation complication prediction system

Abstract

The present application relates to the technical field of respiratory monitoring, in particular to an ERCP sedation complication prediction system based on AI. The present application determines a reference respiratory conduction time delay and a reference amplitude ratio based on the time sequence correlation and amplitude response characteristics of the electroencephalogram respiratory modulation value and the pulse respiratory modulation value in the body position stable period; determines a comprehensive pairing cost based on the deviation degree of the time difference between any pulse wave peak time and any electroencephalogram wave peak time in the monitoring period relative to the reference respiratory conduction time delay and the deviation degree of the amplitude ratio relative to the reference amplitude ratio, and constructs a bipartite graph model containing the electroencephalogram wave peak time and the pulse wave peak time to determine an optimal physiological pairing set; and evaluates the respiratory state of the subject according to the number of electroencephalogram wave peak times not contained in the optimal physiological pairing set and the comprehensive pairing cost in the optimal physiological pairing set. The present application improves the respiratory state monitoring effect caused by nonlinear distortion under the prone position by establishing a body position reference and performing driving response causal matching.

Description

Technical Field

[0001] This invention relates to the field of respiratory monitoring technology, and more specifically to an AI-based ERCP sedation complication prediction system. Background Technology

[0002] Endoscopic retrograde cholangiopancreatography (ERCP) is a key technique for diagnosing and treating biliary and pancreatic diseases. This procedure typically requires the patient to be in a prone or semi-prone position and to use sedatives (such as remimazolam) to suppress body movement. In this specific position, monitoring respiratory complications presents significant challenges: the prone position causes compression of the chest and abdomen, potentially altering chest wall compliance and diaphragmatic movement; simultaneously, sedatives may induce central respiratory depression or upper airway collapse, both of which increase the risk of hypoventilation.

[0003] Existing patent document CN119132578A discloses a monitoring and control system and method for anesthesia intubation, specifically assessing respiratory and circulatory function by analyzing the subject's pathological data and blood gas monitoring data. However, in the prone position scenario of ERCP, the chest and abdominal compression caused by the prone position physically alters the transmission path of respiratory motion from drive to response, resulting in a nonlinear time drift and amplitude attenuation in the transmission relationship between central respiratory drive signals and peripheral ventilation response signals. This nonlinear distortion introduced by body position makes it difficult for conventional algorithms to accurately establish the correspondence between drive and response, leading to inaccurate respiratory status monitoring. Summary of the Invention

[0004] To address the technical problem of nonlinear distortion of physiological signals caused by the prone position in ERCP, resulting in a mismatch between respiratory drive and response and thus poor respiratory status monitoring, the present invention aims to provide an AI-based ERCP sedation complication prediction system. The specific technical solution adopted is as follows:

[0005] This invention proposes an AI-based ERCP sedation complication prediction system, the system comprising:

[0006] The data acquisition module is used to acquire the EEG respiratory modulation value and pulse respiratory modulation value of the subject at each moment during the ERCP operation period, which includes the body position stabilization period and the monitoring period.

[0007] The baseline calibration module is used to determine the baseline respiratory conduction delay to baseline amplitude ratio based on the temporal correlation and amplitude responsiveness of EEG respiratory modulation values ​​and pulse respiratory modulation values ​​during a period of stable body position.

[0008] The cost analysis module is used to extract the peak time of EEG and the peak time of pulse wave in the preset analysis period for each moment in the monitoring period. Based on the deviation of the time difference between any pulse wave peak time and any EEG peak time relative to the baseline respiratory conduction delay and the deviation of the amplitude ratio relative to the baseline amplitude ratio, the comprehensive pairing cost between the corresponding two moments is determined.

[0009] The physiological matching module is used to construct a bipartite graph model containing the peak times of EEG and pulse waves within the preset analysis period, and to perform global optimization matching based on the comprehensive matching cost to determine the optimal physiological matching set for each moment within the monitoring period.

[0010] The status assessment module is used to assess the subject's respiratory status at each moment based on the number of EEG peak moments not included in the optimal physiological pairing set within the preset analysis period and the corresponding comprehensive pairing cost of the time pairs in the optimal physiological pairing set.

[0011] Further, determining the ratio of the baseline respiratory conduction delay to the baseline amplitude includes:

[0012] The EEG respiratory modulation values ​​and pulse respiratory modulation values ​​at all times during the period of stable body position were arranged in chronological order to obtain the EEG respiratory sequence and pulse respiratory sequence.

[0013] Calculate the cross-correlation function values ​​of the EEG respiratory sequence and pulse respiratory sequence under different time delays within a preset physiological reasonable range, and select the time delay corresponding to the largest cross-correlation function value as the benchmark respiratory conduction time delay;

[0014] The EEG respiratory modulation value at time i within the stable body position period and the value at time t The pulse-respiration modulation values ​​at each moment constitute a data pair; where the value of i ranges from 1 to 1. , The baseline respiratory conduction delay corresponds to the number of time points, and N is the total number of time points within the stable body temperature period.

[0015] Linear fitting is performed on all data pairs corresponding to the stable body position period. It is determined whether the slope of the obtained fitted line is greater than zero. If it is, the slope is used as the reference amplitude ratio; otherwise, a preset minimum positive number is used as the reference amplitude ratio.

[0016] Furthermore, the method for determining the comprehensive pairing cost between two corresponding time points includes:

[0017] Peak values ​​of EEG respiratory modulation and pulse respiratory modulation during the period of stable body position were detected, and the times corresponding to the obtained peak values ​​were recorded as the EEG reference peak time and the pulse reference peak time, respectively.

[0018] Select any EEG baseline peak time as the example time. Calculate the time difference between the sum of the example time and the baseline respiratory conduction delay and all pulse baseline peak times during the stable body position period. Select the pulse baseline peak time corresponding to the smallest time difference as the pulse matching time of the example time. Calculate the standard deviation of the time difference between all EEG baseline peak times and their pulse matching times during the stable body position period as the conduction delay dispersion.

[0019] Calculate the ratio of the absolute value of the pulse-breath modulation value at the pulse-matching time of each EEG reference peak to the absolute value of the EEG-breath modulation value at that EEG reference peak. Use the standard deviation of the logarithmic difference between the ratio corresponding to each EEG reference peak and the reference amplitude ratio during the body position stability period as the amplitude ratio dispersion.

[0020] Based on the deviation of the time difference between any EEG peak moment and any pulse peak moment within the preset analysis period within each monitoring period relative to the conduction delay dispersion, and the deviation of the amplitude ratio relative to the benchmark amplitude ratio, the comprehensive pairing cost between the corresponding two moments is determined.

[0021] Furthermore, the formula for calculating the comprehensive pairing cost is as follows:

[0022] ;in, The comprehensive pairing cost between the i-th EEG peak time and the j-th pulse wave peak time within the preset analysis period for each moment in the monitoring period; To monitor the i-th EEG peak moment within a preset analysis period for each moment in the monitoring period; The absolute value of the EEG respiratory modulation value at the i-th EEG peak time within the preset analysis period is used to monitor each time point within the monitoring period. To determine the j-th pulse peak moment within a preset analysis period for each moment in the monitoring period; The absolute value of pulse-respiration modulation at the j-th pulse peak moment within the preset analysis period is used to monitor each moment within the monitoring period. The number of time points corresponding to the baseline respiratory conduction delay; The baseline amplitude ratio; The propagation delay dispersion; The amplitude ratio dispersion; is a pre-defined minimum positive number; ln is an exponential function with the natural constant as the base;

[0023] If the absolute difference between the time difference between the i-th EEG peak and the j-th pulse wave peak and the baseline respiratory conduction delay within the preset analysis period at each moment in the monitoring period is greater than the preset physiologically reasonable threshold, or if the i-th EEG peak is greater than the j-th pulse wave peak, then... Set to the preset maximum penalty value.

[0024] Furthermore, determining the optimal physiological pairing set for each moment within the monitoring period includes:

[0025] For each moment within the monitoring period, the sets of peak EEG times and peak pulse times within the preset analysis period for each moment are respectively denoted as the EEG time set and the pulse time set.

[0026] The absolute difference in the number of elements between the EEG time set and the pulse time set is recorded as the number to be compensated; virtual filling nodes of the number to be compensated are added to the time set with the smallest number of elements, and the corresponding time set is updated; the peak time of EEG and the peak time of pulse are recorded as the physiological peak time.

[0027] A bipartite graph model is constructed based on the updated EEG time set and pulse time set. In the model, the weight of the edge connecting two physiological peak times is set as the comprehensive pairing cost between the two times, and the weight of the edge connecting the physiological peak times and the virtual filling node is set as a preset maximum penalty value.

[0028] The minimum weight perfect matching of the bipartite graph model is solved using the Hungarian algorithm to obtain an initial physiological matching set consisting of several time pairs; wherein, each time pair contains one element from the updated EEG time set and one element from the updated pulse time set.

[0029] Remove time pairs containing the virtual filler nodes from the initial physiological matching set, and the remaining time pairs constitute the optimal physiological matching set.

[0030] Furthermore, the assessment of the subject's respiratory status at each moment includes

[0031] For each moment within the monitoring period, the percentage of EEG peak moments not included in the optimal physiological pairing set within the preset analysis period of each moment is taken as the circuit breaker rate.

[0032] The congestion level is calculated as the arithmetic mean of the comprehensive pairing costs for all time pairs in the optimal physiological pairing set at each time moment.

[0033] Based on the circuit breaker rate and the congestion level, the respiratory status of the subjects at each time point is assessed.

[0034] Furthermore, the assessment of the subject's respiratory status at each moment based on the circuit breaker rate and the congestion level includes:

[0035] If the number of peak EEG moments within the preset analysis period at each moment is less than the preset central inhibition threshold, then the subject's respiratory state at that moment is central inhibition.

[0036] If the number of peak EEG moments within a preset analysis period at each moment is greater than or equal to a preset central inhibition threshold, and the circuit breaker rate is greater than a preset circuit breaker threshold, then the subject's respiratory state at that moment is complete obstruction.

[0037] If the number of peak EEG moments in the preset analysis period at each moment is greater than or equal to the preset central inhibition threshold, the circuit breaker rate is less than or equal to the preset circuit breaker threshold, and the congestion degree is greater than the preset congestion threshold, then the subject's respiratory state at that moment is compensated obstruction.

[0038] If the number of peak EEG moments in the preset analysis period at each moment is greater than or equal to the preset central inhibition threshold, the circuit breaker rate is less than or equal to the preset circuit breaker threshold, and the congestion degree is less than or equal to the preset congestion threshold, then the subject's breathing state at each moment is normal breathing.

[0039] Furthermore, the extraction of the peak times of electroencephalogram (EEG) and pulse waves within a preset analysis period for each moment during the monitoring period includes:

[0040] The brainwave respiratory modulation value and the pulse respiratory modulation value are recorded as physiological modulation values;

[0041] For each moment within the monitoring period, perform a local maximum search for the same physiological modulation value at all moments within the preset analysis period, and record the moment corresponding to the maximum value as the analysis moment for each physiological modulation value;

[0042] The mean and standard deviation of each physiological modulation value are calculated for all times within the preset analysis period. The sum of the mean and K times the standard deviation is used as the judgment threshold for each physiological modulation value; where K is a preset constant.

[0043] For each physiological modulation value, the analysis time when the physiological modulation value is greater than the judgment threshold is selected as the effective peak time; the effective peak times corresponding to the EEG respiratory modulation value and the pulse respiratory modulation value are recorded as the EEG peak time and the pulse peak time, respectively.

[0044] Furthermore, the least squares method was used to perform linear fitting on all data pairs corresponding to the period of stable body position.

[0045] Furthermore, K is set to 1.

[0046] The present invention has the following beneficial effects:

[0047] In this embodiment of the invention, by measuring the ratio of baseline respiratory conduction delay to baseline amplitude during a stable body position, an individualized physiological transmission baseline adapted to the current body position is established for the subject. This effectively eliminates interference from non-pathological factors such as chest and abdominal pressure caused by the prone position on respiratory signal transmission. Based on the deviation of the time difference and amplitude ratio of the two peak times relative to the baseline value, a comprehensive pairing cost is constructed from the temporal and spatial dimensions to quantify the causal relationship between respiratory drive and response. Global optimization matching is performed based on the comprehensive pairing cost to purify the drive-response pairings that conform to physiological laws from signals with noise and distortion, forming the optimal physiological pairing set. By comprehensively considering the number of unmatched EEG peaks and the pairing cost, the respiratory status is accurately assessed from two dimensions: signal connectivity and transmission quality. Through the above operations, this solution can solve the problem of respiratory drive and response mismatch caused by nonlinear signal distortion in the prone position during ERCP, significantly improving the accuracy, robustness, and clinical reliability of respiratory status monitoring, and effectively reducing the false alarm rate. Attached Figure Description

[0048] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0049] Figure 1 A system structure diagram of an AI-based ERCP sedation complication prediction system provided in an embodiment of the present invention;

[0050] Figure 2 This is a schematic diagram of a computer device for predicting ERCP sedation complications based on AI, provided as an embodiment of the present invention. Detailed Implementation

[0051] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of an AI-based ERCP sedation complication prediction system proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0052] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0053] The following description, in conjunction with the accompanying drawings, details the specific scheme of the AI-based ERCP sedation complication prediction system provided by this invention.

[0054] Example 1:

[0055] Please see Figure 1 The diagram illustrates a system block diagram of an AI-based ERCP sedation complication prediction system according to an embodiment of the present invention. The system includes: a data acquisition module 110, a benchmark calibration module 120, a cost analysis module 130, a physiological matching module 140, and a status assessment module 150.

[0056] The data acquisition module 110 is used to acquire the EEG respiratory modulation value and pulse respiratory modulation value of the subject at each moment during the ERCP operation period, which includes the period of body position stabilization and the monitoring period.

[0057] Specifically, the electroencephalogram (EEG) signals containing respiratory center driving information are collected by electrodes attached to the subject's forehead or mastoid region, and pulse wave signals reflecting changes in peripheral blood perfusion are collected by transmissive or reflective photoelectric oxygenation probes clipped to the subject's fingers or earlobes. It is necessary to ensure that the two signals are collected synchronously at the same sampling rate, so as to obtain the subject's EEG and pulse values ​​at various times during the ERCP procedure.

[0058] The surgical period refers to the entire process from the completion of anesthesia induction and positioning of the patient until the patient's recovery after surgery. The postural stability period refers to the time window during which physiological signals and vital signs reach a stable state after the patient completes the prone position roll. Specifically, after the patient maintains the prone position, the signal quality index and heart rate variability coefficient of the pulse-respiratory modulation values ​​are continuously monitored. The signal quality index refers to the skewness of the pulse-respiratory modulation values ​​within a preset time window. The preset time window is a sliding time window, with its length set to include at least one complete respiratory cycle, for example, 5 seconds, and slides on the time axis in preset steps (e.g., 1 second) to calculate the statistical characteristics of the data within the current window in real time. When the signal quality index is detected to be continuously higher than a preset quality threshold (e.g., 0.8) for a preset stable duration (e.g., 30 seconds) and the heart rate variability coefficient is lower than a preset stable threshold (e.g., 10%), this time period meeting the conditions is considered the postural stability period. The aim is to ensure that data used for baseline calibration are collected when the signal is clear and the subject's physiological state is stable, avoiding deviations in baseline parameter calibration due to motion artifacts or physiological stress responses during the initial stage of turning over. The monitoring period refers to the entire surgical procedure following the period of postural stability. To ensure the validity of the baseline parameters, the subject must maintain the same position as during the postural stability period during the monitoring period, usually in a prone position.

[0059] The EEG and pulse values ​​at all times during the surgical period were arranged chronologically to obtain EEG and pulse value sequences, respectively, and then denoised using a moving average method. Subsequently, a first bandpass filter with a passband of 0.1 to 1 Hz was used to filter out the heartbeat component and baseline drift in the pulse value sequence, retaining the low-frequency pulse waveform reflecting respiratory modulation. Although the EEG signal mainly reflects cortical electrical activity, respiratory movements generate modulation components in the EEG signal synchronized with the respiratory frequency through scalp micromovements, hemodynamic coupling, and artifact introduction. This embodiment uses a bandpass filter to extract this component as a central respiratory reference signal. Specifically, a second bandpass filter of 30 to 100 Hz is used to extract the high-frequency components in the EEG value sequence, and the instantaneous amplitude envelope is obtained through Hilbert transform. Then, the signal is demodulated using the same 0.1 to 1 Hz bandpass filter to obtain a respiratory drive waveform reflecting the central respiratory rhythm. Finally, the corresponding values ​​on the preserved low-frequency pulse waveform and respiratory drive waveform at each moment during the surgical period are recorded as the EEG respiratory modulation value and pulse respiratory modulation value at that moment.

[0060] In one implementation of this invention, the sampling rate of both the electroencephalogram (EEG) signal and the pulse wave signal is set to 250 Hz. The implementer can set this according to the specific circumstances.

[0061] The reference calibration module 120 is used to determine the reference respiratory conduction delay to reference amplitude ratio based on the temporal correlation and amplitude responsiveness of EEG respiratory modulation values ​​and pulse respiratory modulation values ​​during a period of stable body position.

[0062] During ERCP, the patient needs to change from a supine to a prone position. This positional change physically alters the mechanical characteristics of respiratory motion transmission from the diaphragm to peripheral blood flow due to chest and abdominal pressure, leading to non-pathological deviations in signal transmission delay and amplitude attenuation. Therefore, this step, during a period of stable positioning, utilizes the temporal correlation and amplitude response characteristics of EEG respiratory modulation values ​​and pulse respiratory modulation values ​​to determine the ratio of baseline respiratory conduction delay to baseline amplitude. This aims to obtain the patient's physiological state in a specific position. This effectively eliminates interference from positional factors and individual differences on the signal transmission model, ensuring that subsequent data can accurately identify pathological deviations caused by sedation drugs or airway obstruction, using the normal transmission state in the current position as a benchmark, thereby significantly reducing the false alarm rate.

[0063] The cost analysis module 130 is used to extract the peak time of EEG and the peak time of pulse wave in the preset analysis period for each moment in the monitoring period. Based on the deviation of the time difference between any pulse wave peak time and any EEG peak time relative to the benchmark respiratory conduction delay and the deviation of the amplitude ratio relative to the benchmark amplitude ratio, the comprehensive pairing cost between the corresponding two moments is determined.

[0064] Since the peak characteristics are not easily affected by baseline drift and high-frequency noise, they can stably and accurately characterize the complete "central drive and peripheral response" respiratory event, and are the best plotting points for establishing causal matching relationships. Therefore, the peak times of EEG and pulse waves during the monitoring period are selected.

[0065] In prone ERCP surgery, the time difference and amplitude ratio between the peak pulse wave and the peak EEG wave reflect the physical delay of neural command transmission to the periphery and the mechanical efficiency of drive conversion into ventilation, respectively. The deviation of the time difference from the baseline respiratory conduction delay and the deviation of the amplitude ratio from the baseline amplitude ratio, respectively, present the causal relationship between respiratory drive and response from the perspectives of transmission delay deviation in the time dimension and amplitude response deviation in the spatial dimension. The comprehensive pairing cost obtained by combining the two can quantify the possibility that the peak pulse wave and the peak EEG wave belong to the same respiratory action, and can more reliably identify true pairings that conform to physiological laws, avoid misjudgments caused by local waveform distortion, and thus significantly improve the robustness of the system in complex positions.

[0066] In one implementation of this invention, the 30 seconds preceding each moment within the monitoring period are designated as the preset analysis period for that moment. The duration of the preset analysis period can be set according to specific circumstances. The respiratory status of the subject is not analyzed during the first 30 seconds of the monitoring period.

[0067] The physiological matching module 140 is used to construct a bipartite graph model containing the peak times of EEG and pulse waves within a preset analysis period. It performs global optimization matching based on the comprehensive matching cost to determine the optimal physiological matching set for each moment within the monitoring period.

[0068] Due to the prone position during ERCP and the effects of sedation, respiratory physiological signals are often accompanied by nonlinear drift and random artifacts. Simple local thresholding or nearest-neighbor matching often fails to distinguish between genuine driver-response relationships and random noise. This step transforms the attribution problem of respiratory events into a minimum-weight matching problem in graph theory by constructing a bipartite graph model and performing global optimization matching based on the comprehensive pairing cost. This forces all potential driver-response pairings to compete for the goal of minimizing the total deviation cost. This mechanism eliminates erroneous pairings that result in a surge in overall cost through local matching, preserving the most physiologically reasonable true causal transmission relationships to form the optimal physiological pairing set, thus improving the high specificity of purifying true respiratory events under complex interference environments.

[0069] The status assessment module 150 is used to assess the respiratory status of the subject at each moment based on the number of EEG peak moments not included in the optimal physiological pairing set within the preset analysis period and the corresponding comprehensive pairing cost of the time pairs in the optimal physiological pairing set.

[0070] The number of unmatched EEG peaks reflects the degree to which central driving signals are completely lost during transmission, while the overall pairing cost reflects the time delay and amplitude attenuation experienced by the signal during transmission. This allows the system to penetrate the apparent signal chaos and accurately identify the true respiratory pathology state masked by prone position distortion from both connectivity and transmission quality dimensions, thereby improving the monitoring effect of respiratory status in complex positions.

[0071] Preferably, in some possible implementations of the present invention, the method for obtaining the ratio of the reference respiratory conduction delay to the reference amplitude includes: arranging the EEG respiratory modulation values ​​and pulse respiratory modulation values ​​at all times within the stable body position period in chronological order to obtain the EEG respiratory sequence and the pulse respiratory sequence; calculating the cross-correlation function values ​​of the EEG respiratory sequence and the pulse respiratory sequence at different delays within a preset physiologically reasonable interval, and selecting the delay corresponding to the largest cross-correlation function value as the reference respiratory conduction delay; and using the ratio of the EEG respiratory modulation value at the i-th time moment within the stable body position period to the reference amplitude value at the i-th time moment. The pulse-respiration modulation values ​​at each moment constitute a data pair; where the value of i ranges from 1 to 1. , The baseline respiratory conduction delay corresponds to the number of time points, and N is the total number of time points within the body temperature stabilization period. Linear fitting is performed on all data pairs corresponding to the body position stabilization period, and it is determined whether the slope of the obtained fitted line is greater than zero. If it is, the slope is used as the baseline amplitude ratio; otherwise, a preset minimum positive number is used as the baseline amplitude ratio.

[0072] It should be noted that cross-correlation analysis was used to determine the inherent physical delay of respiratory drive conduction. The cross-correlation function value reflects the waveform similarity between the EEG respiratory sequence and the pulse respiratory sequence at each time delay. The larger the value, the more consistent the trend of the two sequences at that time delay. The baseline respiratory conduction delay reflects the actual physical lag time of signal transmission from the central nervous system to the peripheral blood vessels under the current prone position compression.

[0073] Because there is an inherent physical lag in the transmission of human nerve signals to peripheral hemodynamic changes, namely the baseline respiratory conduction delay, the EEG respiratory modulation value and the pulse respiratory modulation value collected at the same moment do not correspond to the same respiratory action. Therefore, the pulse respiratory sequence is traced back Tdelay moments on the time axis, aligning the respiratory drive command issued at moment i with the actual ventilation response generated at moment i+Tdelay, forming a data pair. This eliminates the interference of the time factor and truly reflects the effective ventilation response amplitude that a unit intensity of central drive can stimulate. Linear fitting is performed on all data pairs corresponding to the postural stability period. Specifically, a two-dimensional coordinate system is constructed with the EEG respiratory modulation value as the horizontal axis and the pulse respiratory modulation value as the vertical axis. All data pairs corresponding to the postural stability period are marked in the two-dimensional coordinate system to obtain the coordinate points of the corresponding data pairs. The least squares method is then used to perform linear fitting on the coordinate points in the two-dimensional coordinate system.

[0074] Under normal respiratory mechanisms, an increase in central respiratory intensity should lead to diaphragmatic contraction and increased pulmonary ventilation, which in turn leads to a corresponding increase in the amplitude of peripheral hemodynamic fluctuations. This indicates a positive causal relationship between central respiratory rhythm and peripheral ventilation response. If the slope is greater than zero, it indicates that enhanced EEG-driven signals trigger enhanced pulse response signals, consistent with normal respiratory mechanisms, and the slope is used as the baseline amplitude ratio. Conversely, if the slope is less than zero, it indicates severe signal abnormalities or non-physiological interference, violating normal respiratory mechanisms. To ensure the mathematical validity and system stability of subsequent calculations, a preset minimum positive number is used as the baseline amplitude ratio. In this embodiment, the preset minimum positive number is set to 0.1. The baseline amplitude ratio reflects the ventilation efficiency of the entire respiratory pathway in the subject's current position and baseline physiological state.

[0075] In one implementation of this invention, the preset physiological reasonable range is set based on the physiological statistical laws of human nerve conduction and hemodynamic response. According to clinical monitoring data, the physical delay of respiratory drive conduction to the peripheral pulse wave in the adult population typically ranges from 0.5 seconds to 2.5 seconds. To ensure that the preset physiological reasonable range effectively covers the individualized physiological delay characteristics of the vast majority of subjects, this embodiment sets this range as follows: Starting from the lower limit of the interval, traverse the interval with a step size of 0.1 seconds to obtain multiple time delays.

[0076] It is important to note that to ensure the reliability of the baseline parameters, the fitting results must be validated. If the goodness of fit of the linear fit is less than the preset confidence threshold (e.g., 0.6), or the slope of the fitted line is not greater than 0, it means that the linear correlation between EEG drive and pulse response is weak, a stable transmission model cannot be established, or the drive enhancement fails to induce response enhancement or even shows an inverse change, violating the basic physiological causal law that respiratory drive induces positive ventilation response, and the baseline calibration fails. In this case, a preset population experience value (e.g., baseline respiratory conduction delay set to 0.8 seconds, baseline amplitude ratio set to 1) will be used as a temporary baseline parameter, or the monitoring time during the postural stabilization period will be extended until a calibration result that meets the validity validation requirements is obtained. Preferably, in some possible implementations of the embodiments of the present invention, the method for obtaining the peak time of the EEG and the peak time of the pulse wave includes: recording the EEG respiratory modulation value and the pulse respiratory modulation value as physiological modulation values; performing a local maximum search on the same physiological modulation value at all times within a preset analysis period for each time within the monitoring period, and recording the time corresponding to the maximum value as the analysis time of each physiological modulation value; calculating the mean and standard deviation of each physiological modulation value at all times within the preset analysis period, and using the sum of the mean and K times the standard deviation as the judgment threshold for each physiological modulation value; where K is a preset constant; for each physiological modulation value, selecting the analysis time where the physiological modulation value is greater than the judgment threshold and recording it as the effective peak time; and recording the effective peak times corresponding to the EEG respiratory modulation value and the pulse respiratory modulation value as the EEG peak time and the pulse peak time, respectively.

[0077] It should be noted that the patient's basal metabolic rate and external interference change in real time during ERCP. In this embodiment, a judgment threshold that floats with the signal baseline can be dynamically constructed to filter out most background random noise and retain only effective physiological fluctuation events that deviate significantly from the baseline. This ensures the sensitivity of peak extraction while suppressing false positive interference to the greatest extent, laying the foundation for accurate matching in the future.

[0078] In one implementation of this invention, a preset constant K is set to 1, such that the judgment threshold corresponds to a 68.3% probability interval under a normal distribution, which can eliminate most random background noise fluctuations.

[0079] Preferably, in some possible implementations of the embodiments of the present invention, the method for obtaining the comprehensive pairing cost includes: performing peak detection on the EEG respiratory modulation value and the pulse respiratory modulation value during the stable body position period, and recording the times corresponding to the obtained peaks as the EEG reference peak time and the pulse reference peak time in sequence; randomly selecting one EEG reference peak time as the example time, calculating the time difference between the sum of the example time and the reference respiratory conduction delay and all pulse reference peak times during the stable body position period, and selecting the pulse reference peak time corresponding to the smallest time difference as the pulse matching time of the example time; calculating the pulse matching time of all EEG reference peak times during the stable body position period. The standard deviation of the time difference between the matching moments is used as the conduction delay dispersion. The ratio of the absolute value of the pulse-respiratory modulation value at the pulse matching moment to the absolute value of the EEG-respiratory modulation value at each EEG reference peak moment is calculated. The standard deviation of the logarithmic difference between the ratio corresponding to each EEG reference peak moment and the reference amplitude ratio within the stable body position period is used as the amplitude ratio dispersion. Based on the deviation of the time difference between any EEG peak moment and any pulse peak moment within the preset analysis period within the monitoring period, relative to the conduction delay dispersion, and the deviation of the amplitude ratio relative to the reference amplitude ratio, the comprehensive pairing cost between the corresponding two moments is determined. The expression for the logarithmic difference is... , This represents the absolute value of the pulse-respiratory modulation at the nth EEG baseline peak within the territory during a period of postural stability. The absolute value of the EEG respiratory modulation value at the nth EEG reference peak time within the territory during the period of stable body position; This is the baseline amplitude ratio. It should be noted that the brain exhibits weak electrical activity even under deep sedation. It will not be zero.

[0080] It should be noted that due to the inherent time lag in the conversion of respiratory drive signals into peripheral blood flow responses via neural conduction and mechanical ventilation, the sum of the example time and the baseline respiratory conduction delay represents the ideal time when the peripheral pulse should exhibit a ventilatory response peak for the same respiratory action as the example time. The pulse baseline peak time closest to this sum represents the actual time point in that respiratory action observed to be the time when the actual ventilatory response triggered by the EEG drive at the example time occurs, thus obtaining the pulse matching time. The time difference between the EEG baseline peak time and the pulse matching time represents the actual transmission delay from the brain issuing the command to the finger producing a pulse pulsation for each respiratory action. The conduction delay dispersion reflects the degree of fluctuation in the respiratory signal transmission time of the subject in the current position. The ratio of the absolute value of the pulse-respiratory modulation value at the pulse matching time of the EEG baseline peak time to the absolute value of the EEG-respiratory modulation value at the EEG baseline peak time represents the actual ventilation efficiency for each respiratory action. The amplitude ratio dispersion reflects the degree of fluctuation in the respiratory ventilation efficiency of the subject in the current position. The two dispersions serve as normalization factors for analyzing the overall pairing cost.

[0081] It is important to note that the method for obtaining the pulse matching time of all EEG reference peak times during the stable body temperature period is the same as that for the example time. In this embodiment of the invention, if the time difference between the sum of the example time and the reference respiratory conduction delay and all pulse reference peak times during the stable body position period is greater than or equal to a preset physiological limit (e.g., 1 second), it indicates that there may be transient signal loss, severe artifact interference, or the EEG peak is spurious noise. In this case, the pairing relationship is not valid, and it is also considered that there is no pulse matching time for the example time. EEG reference peak times without pulse matching times are not included in the calculation process of conduction delay dispersion and amplitude ratio dispersion.

[0082] In one specific implementation of this invention, the comprehensive pairing cost is expressed by the formula:

[0083]

[0084] in, The comprehensive pairing cost between the i-th EEG peak time and the j-th pulse wave peak time within the preset analysis period for each moment in the monitoring period; To monitor the i-th EEG peak moment within a preset analysis period for each moment in the monitoring period; The absolute value of the EEG respiratory modulation value at the i-th EEG peak time within the preset analysis period is used to monitor each time point within the monitoring period. To determine the j-th pulse peak moment within a preset analysis period for each moment in the monitoring period; The absolute value of pulse-respiration modulation at the j-th pulse peak moment within the preset analysis period is used to monitor each moment within the monitoring period. The number of time points corresponding to the baseline respiratory conduction delay; The baseline amplitude ratio; This represents the dispersion of the conduction delay. The amplitude ratio dispersion; The value is a preset, extremely small positive number to prevent the denominator from being zero or the logarithmic domain from being invalid. In this embodiment, it is set to 0.1% of the mean absolute value of the EEG respiratory modulation values ​​during the period of stable posture; ln is an exponential function with the natural constant as the base. It should be noted that... Reflects the actual transmission delay of nerve signals in the human body to peripheral blood vessels The degree of deviation from the baseline respiratory conduction delay, Reflects the ratio of actual ventilation efficiency per breath to the baseline amplitude. The degree of deviation. The comprehensive pairing cost comprehensively measures the degree of deviation between the i-th EEG peak and the j-th pulse peak in both time and amplitude intensity dimensions relative to the subject's individual baseline state. The larger the value, the less likely that the two peaks were generated by the same breathing action.

[0085] To eliminate pairings that violate physiological laws, a mandatory constraint is introduced: if the absolute difference between the time difference between the i-th EEG peak and the j-th pulse peak and the baseline respiratory conduction delay is greater than a preset physiologically reasonable threshold, it indicates possible instantaneous signal loss, severe artifact interference, or the EEG peak being spurious noise; or if the i-th EEG peak is greater than the j-th pulse peak, it indicates that the two times violate the causal law that the central respiratory drive time must be earlier than the peripheral ventilation response time, meaning that the two times cannot physiologically correspond to the same respiratory action. The maximum penalty value is set to completely block the pairing of the two in subsequent analysis, ensuring that the matching results strictly follow physiological logic.

[0086] In this embodiment of the invention, a preset physiologically reasonable threshold is used to limit the maximum permissible time drift range between respiratory drive and response. Its setting is typically based on the subject's time dispersion in a baseline state, and can be set to three times the conduction delay dispersion. However, to prevent the threshold from being too narrow due to excessively narrow conduction delay dispersion in extreme cases where the baseline state is too stable, the maximum value between a preset empirical minimum value (e.g., 0.5 seconds) and three times the conduction delay dispersion is selected as the preset physiologically reasonable threshold to ensure that normal physiological rhythm fluctuations are tolerated. In this embodiment, the preset maximum penalty value should be much greater than the comprehensive pairing cost of the subsequent Hungarian algorithm's priority selection of all possible true physiological peak time pairs. The preset maximum penalty value can be set to... .

[0087] Preferably, in some possible implementations of the embodiments of the present invention, the method for obtaining the optimal physiological pairing set includes: for each moment within the monitoring period, the sets consisting of the peak EEG time and the peak pulse time within the preset analysis period for each moment are respectively denoted as the EEG time set and the pulse time set; the absolute difference in the number of elements between the EEG time set and the pulse time set is denoted as the number to be compensated; virtual filling nodes for the number to be compensated are added to the time set with the smallest number of elements, and the corresponding time set is updated; the peak EEG time and the peak pulse time are denoted as the physiological peak time; based on the updated EEG time set and the pulse time set... A bipartite graph model is constructed using the pulse time set. The weight of the edge connecting two physiological peak times is set as the comprehensive pairing cost between the two times, and the weight of the edge connecting the physiological peak times and the virtual filling node is set with a preset maximum penalty value. The minimum weight perfect match of the bipartite graph model is solved using the Hungarian algorithm to obtain an initial physiological matching set consisting of several time pairs. Each time pair contains one element from the updated EEG time set and one element from the updated pulse time set. Time pairs containing virtual filling nodes are removed from the initial physiological matching set, and the remaining time pairs constitute the optimal physiological matching set.

[0088] It should be noted that during actual monitoring, due to the asynchrony of physiological signals (such as the inconsistency between heart rate and respiratory rate) and noise interference, the number of EEG peak moments and the number of pulse wave peak moments within the monitoring period are usually not equal. It is necessary to add virtual fill nodes to the time set with the smallest number of elements to complete it, thereby constructing a bipartite graph model with balanced elements. The virtual fill node is a virtual mathematical placeholder introduced to solve the graph matching problem. The time set without virtual fill nodes is directly used as the updated time set.

[0089] To transform the complex temporal causal relationships and amplitude response characteristics of physiological signals into a solvable mathematical optimization objective, the weight of the edge connecting two physiological peak moments is set as the comprehensive pairing cost between the two moments. This aims to leverage the optimization characteristic of the bipartite graph matching algorithm—minimizing the sum of global weights—to select the driver-response pairings that best conform to the subject's baseline physiological transmission patterns, thereby effectively overcoming matching ambiguities caused by nonlinear signal drift in the prone position. Specifically, time pairs containing virtual filler nodes represent peak moments where a reasonable causal relationship cannot be established. The weight of the corresponding edge is set to a preset maximum penalty value, eliminating all time pairs containing virtual filler nodes and retaining only the optimal physiological pairing set that conforms to physiological causality and minimizes the total cost.

[0090] Preferably, in some possible implementations of the embodiments of the present invention, the method for assessing respiratory status includes: for each moment within the monitoring period, taking the proportion of EEG peak moments not included in the optimal physiological pairing set of that moment within the preset analysis period of that moment as the circuit breaker rate; calculating the arithmetic mean of the comprehensive pairing costs of all moments in the optimal physiological pairing set of each moment as the congestion level; and assessing the respiratory status of the subject at each moment based on the circuit breaker rate and the congestion level.

[0091] It should be noted that the disconnection rate characterizes the connectivity of the respiratory pathway. The higher the value, the more respiratory commands issued by the brain are lost during transmission without producing an effective peripheral response, i.e., the poorer the connectivity of the respiratory pathway. The congestion rate characterizes the transmission resistance of the respiratory pathway. The higher the value, the more significant the time delay or amplitude attenuation compared to the baseline state, even if the respiratory command is successfully transmitted and a pulse response is generated.

[0092] It should be noted that if there is no peak EEG moment in the preset analysis period at any time during the monitoring period, the circuit breaker rate will be set to zero; if the optimal physiological matching pair set is empty, the congestion level will be set to zero.

[0093] Central inhibition is characterized by an extremely low central drive frequency, meaning that the number of EEG peaks detected per unit time is sparse; complete obstruction is characterized by a normal central drive frequency, but a large number of drive signals cannot elicit an effective peripheral pulse response; compensatory obstruction is characterized by a significant time delay or amplitude attenuation in the process of central drive response, resulting in low ventilation efficiency; normal breathing is characterized by a patent and functionally normal respiratory pathway.

[0094] If the number of EEG peaks within the preset analysis period is less than the preset central inhibition threshold, it indicates that the subject's respiratory center is firing too few impulses, i.e., the number of EEG peaks is sparse within that period, and the respiratory state is central inhibition. In this case, further analysis of airway patency is unnecessary, and the focus of intervention is on restoring the drive through drug antagonism or physical stimulation.

[0095] If the number of peak EEG moments within the preset analysis period at each moment is greater than or equal to the preset central inhibition threshold, and the circuit breaker rate is greater than the preset circuit breaker threshold, it indicates that the central drive is normal but a large number of drive signals are lost during transmission, which meets the physical characteristics of complete obstruction: "movement without ventilation." In this case, the respiratory state is complete obstruction. The focus of intervention at this time is to physically open the airway. Recommended procedures include: immediately lifting the chin, inserting an oropharyngeal airway, or adjusting the head position.

[0096] If the number of peak EEG moments within the preset analysis period at each time point is greater than or equal to the preset central inhibition threshold, the circuit breaker rate is less than or equal to the preset circuit breaker threshold, and the congestion degree is greater than the preset congestion threshold, it indicates that central drive can induce a peripheral pulse response, but with significant time delay or amplitude attenuation, consistent with the physiological characteristics of compensatory obstruction: "airway narrowing and increased resistance." In this case, the intervention focus is on reducing airway resistance. Recommended procedures include: clearing airway secretions, checking the nasal cannula position, and preparing to open the airway.

[0097] If none of the above conditions are met, that is, if the number of EEG peak moments within the preset analysis period is greater than or equal to the preset central inhibition threshold, the circuit breaker rate is less than or equal to the preset circuit breaker threshold, and the congestion degree is less than or equal to the preset congestion threshold, it indicates that the central drive is sufficient, the pathway is connected and the transmission is smooth, and the breathing state is normal breathing.

[0098] In one implementation of this invention, the preset thresholds can be set based on general clinical physiological standards and large-sample statistical regularities. The preset central inhibition threshold is set according to the medically defined bradyventricular rate of 8 breaths / min. 8 breaths / min can be converted to 8 breaths per 60 seconds, denoted as p. The product of the preset analysis period duration and p is used as the preset central inhibition threshold, where the unit of the preset analysis period duration is seconds. The preset circuit breaker threshold is set based on the signal loss characteristics under complete obstruction, and is set to 0.5, meaning that more than half of the respiratory drives fail to produce an effective response. The preset congestion threshold can be set based on the pairing cost distribution under baseline conditions, and can be set to the 95th percentile of the corresponding comprehensive pairing cost distribution at all times in the optimal physiological pairing set. Clinicians fine-tune the above thresholds according to the specific circumstances of the examinee.

[0099] The respiratory complication prediction results obtained by this method are essentially objective quantitative indicators generated after time-series causal modeling of multimodal physiological signals. These results aim to reveal microscopic transmission abnormalities between central drive and peripheral response that are difficult to capture using conventional monitoring methods. This provides clinicians with auxiliary reference data regarding airway patency and respiratory drive status, thereby improving the efficiency and accuracy of early respiratory complication identification. This protocol does not directly provide a final clinical diagnosis or treatment plan; the final medical diagnosis and intervention decision still needs to be made by a professional physician based on a comprehensive assessment of the patient's real-time vital signs, clinical manifestations, and depth of anesthesia.

[0100] This invention is now complete.

[0101] Example 2:

[0102] Figure 2 This is a schematic diagram of a computer device for predicting ERCP sedation complications based on AI, as provided in one embodiment of the present invention. For example,... Figure 2As shown, the computer device includes: a memory 201, a processor 202, and a computer program 203 stored in the memory 201 and running on the processor 202, wherein when the processor 202 executes the computer program 203, the computer device can execute any of the aforementioned AI-based ERCP sedation complication prediction systems.

[0103] Furthermore, embodiments of this application also protect an apparatus that may include a memory and a processor, wherein the memory stores executable program code, and the processor is used to call and execute the executable program code to execute an AI-based ERCP sedation complication prediction system provided in embodiments of this application.

[0104] This embodiment can divide the device into functional modules based on the above method example. For example, each module can correspond to a separate function, or two or more functions can be integrated into one processing module. The integrated module can be implemented in hardware. It should be noted that the module division in this embodiment is illustrative and only represents one logical functional division. In actual implementation, there may be other division methods.

[0105] It should be understood that the device provided in this embodiment is used to execute the above-described AI-based ERCP sedation complication prediction system, and therefore can achieve the same effect as the above implementation method.

[0106] When using integrated units, the device may include a processing module and a storage module. When applied to a workpiece, the processing module can be used to control and manage the workpiece's operations. The storage module can be used to support the execution of program code by the workpiece.

[0107] The processing module may be a processor or a controller, which can implement or execute various exemplary logic blocks, modules, and circuits as disclosed in this application. The processor may also be a combination of computing functions, such as a combination of one or more microprocessors, a combination of digital signal processing (DSP) and microprocessors, etc., and the storage module may be a memory.

[0108] Example 3:

[0109] This embodiment also provides a computer-readable storage medium storing computer program code. When the computer program code is run on a computer, the computer executes the above-described related method steps to implement the AI-based ERCP sedation complication prediction system provided in the above embodiment.

[0110] In this embodiment, the device and computer-readable storage medium are used to execute the corresponding system provided above. Therefore, the beneficial effects that can be achieved can be referred to the beneficial effects of the corresponding system provided above, and will not be repeated here.

[0111] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0112] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0113] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An AI-based ERCP sedation complication prediction system, characterized in that, The system includes: The data acquisition module is used to acquire the EEG respiratory modulation value and pulse respiratory modulation value of the subject at each moment during the ERCP operation period, which includes the body position stabilization period and the monitoring period. The baseline calibration module is used to determine the baseline respiratory conduction delay to baseline amplitude ratio based on the temporal correlation and amplitude responsiveness of EEG respiratory modulation values ​​and pulse respiratory modulation values ​​during a period of stable body position. The cost analysis module is used to extract the peak time of EEG and the peak time of pulse wave in the preset analysis period for each moment in the monitoring period. Based on the deviation of the time difference between any pulse wave peak time and any EEG peak time relative to the baseline respiratory conduction delay and the deviation of the amplitude ratio relative to the baseline amplitude ratio, the comprehensive pairing cost between the corresponding two moments is determined. The physiological matching module is used to construct a bipartite graph model containing the peak times of EEG and pulse waves within the preset analysis period, and to perform global optimization matching based on the comprehensive matching cost to determine the optimal physiological matching set for each moment within the monitoring period. The status assessment module is used to assess the subject's respiratory status at each moment based on the number of EEG peak moments not included in the optimal physiological pairing set within the preset analysis period and the corresponding comprehensive pairing cost of the time pairs in the optimal physiological pairing set.

2. The AI-based ERCP sedation complication prediction system according to claim 1, characterized in that, The determination of the ratio of the baseline respiratory conduction delay to the baseline amplitude includes: The EEG respiratory modulation values ​​and pulse respiratory modulation values ​​at all times during the period of stable body position were arranged in chronological order to obtain the EEG respiratory sequence and pulse respiratory sequence. Calculate the cross-correlation function values ​​of the EEG respiratory sequence and pulse respiratory sequence under different time delays within a preset physiological reasonable range, and select the time delay corresponding to the largest cross-correlation function value as the benchmark respiratory conduction time delay; The EEG respiratory modulation value at time i within the stable body position period and the value at time t The pulse-respiration modulation values ​​at each moment constitute a data pair; where the value of i ranges from 1 to 1. , The baseline respiratory conduction delay corresponds to the number of time points, and N is the total number of time points within the stable body temperature period. Linear fitting is performed on all data pairs corresponding to the stable body position period. It is determined whether the slope of the obtained fitted line is greater than zero. If it is, the slope is used as the reference amplitude ratio; otherwise, a preset minimum positive number is used as the reference amplitude ratio.

3. The AI-based ERCP sedation complication prediction system according to claim 1, characterized in that, The method for determining the comprehensive pairing cost between two corresponding time points includes: Peak values ​​of EEG respiratory modulation and pulse respiratory modulation during the period of stable body position were detected, and the times corresponding to the obtained peak values ​​were recorded as the EEG reference peak time and the pulse reference peak time, respectively. Select any EEG baseline peak time as the example time. Calculate the time difference between the sum of the example time and the baseline respiratory conduction delay and all pulse baseline peak times during the stable body position period. Select the pulse baseline peak time corresponding to the smallest time difference as the pulse matching time of the example time. Calculate the standard deviation of the time difference between all EEG baseline peak times and their pulse matching times during the stable body position period as the conduction delay dispersion. Calculate the ratio of the absolute value of the pulse-breath modulation value at the pulse-matching time of each EEG reference peak to the absolute value of the EEG-breath modulation value at that EEG reference peak. Use the standard deviation of the logarithmic difference between the ratio corresponding to each EEG reference peak and the reference amplitude ratio during the body position stability period as the amplitude ratio dispersion. Based on the deviation of the time difference between any EEG peak moment and any pulse peak moment within the preset analysis period within each monitoring period relative to the conduction delay dispersion, and the deviation of the amplitude ratio relative to the benchmark amplitude ratio, the comprehensive pairing cost between the corresponding two moments is determined.

4. The AI-based ERCP sedation complication prediction system according to claim 3, characterized in that, The formula for calculating the overall pairing cost is as follows: ;in, The comprehensive pairing cost between the i-th EEG peak time and the j-th pulse wave peak time within the preset analysis period for each moment in the monitoring period; To monitor the i-th EEG peak moment within a preset analysis period for each moment in the monitoring period; The absolute value of the EEG respiratory modulation value at the i-th EEG peak time within the preset analysis period is used to monitor each time point within the monitoring period. To determine the j-th pulse peak moment within a preset analysis period for each moment in the monitoring period; The absolute value of the pulse-respiration modulation value at the j-th pulse peak moment within the preset analysis period is used to monitor each moment within the monitoring period. The number of time points corresponding to the baseline respiratory conduction delay; The baseline amplitude ratio; The propagation delay dispersion; The amplitude ratio dispersion; is a pre-defined minimum positive number; ln is an exponential function with the natural constant as the base; If the absolute difference between the time difference between the i-th EEG peak and the j-th pulse wave peak and the baseline respiratory conduction delay within the preset analysis period at each moment in the monitoring period is greater than the preset physiologically reasonable threshold, or if the i-th EEG peak is greater than the j-th pulse wave peak, then... Set to the preset maximum penalty value.

5. The AI-based ERCP sedation complication prediction system according to claim 1, characterized in that, The determination of the optimal physiological pairing set for each moment within the monitoring period includes: For each moment within the monitoring period, the sets of peak EEG times and peak pulse times within the preset analysis period for each moment are respectively denoted as the EEG time set and the pulse time set. The absolute difference in the number of elements between the EEG time set and the pulse time set is recorded as the number to be compensated; virtual filling nodes of the number to be compensated are added to the time set with the smallest number of elements, and the corresponding time set is updated; the peak time of EEG and the peak time of pulse are recorded as the physiological peak time. A bipartite graph model is constructed based on the updated EEG time set and pulse time set. In the model, the weight of the edge connecting two physiological peak times is set as the comprehensive pairing cost between the two times, and the weight of the edge connecting the physiological peak times and the virtual filling node is set as a preset maximum penalty value. The minimum weight perfect matching of the bipartite graph model is solved using the Hungarian algorithm to obtain an initial physiological matching set consisting of several time pairs; wherein, each time pair contains one element from the updated EEG time set and one element from the updated pulse time set. Remove time pairs containing the virtual filler nodes from the initial physiological matching set, and the remaining time pairs constitute the optimal physiological matching set.

6. The AI-based ERCP sedation complication prediction system according to claim 1, characterized in that, The assessment of the subject's respiratory status at each moment includes For each moment within the monitoring period, the percentage of EEG peak moments not included in the optimal physiological pairing set within the preset analysis period of each moment is taken as the circuit breaker rate. The congestion level is calculated as the arithmetic mean of the comprehensive pairing costs for all time pairs in the optimal physiological pairing set at each time moment. Based on the circuit breaker rate and the congestion level, the respiratory status of the subjects at each time point is assessed.

7. The AI-based ERCP sedation complication prediction system according to claim 6, characterized in that, The assessment of the subject's respiratory status at each moment, based on the circuit breaker rate and the congestion level, includes: If the number of peak EEG moments within the preset analysis period at each moment is less than the preset central inhibition threshold, then the subject's respiratory state at that moment is central inhibition. If the number of peak EEG moments within a preset analysis period at each moment is greater than or equal to a preset central inhibition threshold, and the circuit breaker rate is greater than a preset circuit breaker threshold, then the subject's respiratory state at that moment is complete obstruction. If the number of peak EEG moments in the preset analysis period at each moment is greater than or equal to the preset central inhibition threshold, the circuit breaker rate is less than or equal to the preset circuit breaker threshold, and the congestion degree is greater than the preset congestion threshold, then the subject's respiratory state at that moment is compensated obstruction. If the number of peak EEG moments in the preset analysis period at each moment is greater than or equal to the preset central inhibition threshold, the circuit breaker rate is less than or equal to the preset circuit breaker threshold, and the congestion degree is less than or equal to the preset congestion threshold, then the subject's breathing state at each moment is normal breathing.

8. The AI-based ERCP sedation complication prediction system according to claim 1, characterized in that, The extraction of the peak times of electroencephalogram (EEG) and pulse waves for each time period within the preset analysis time period includes: The brainwave respiratory modulation value and the pulse respiratory modulation value are recorded as physiological modulation values; For each moment within the monitoring period, perform a local maximum search for the same physiological modulation value at all moments within the preset analysis period, and record the moment corresponding to the maximum value as the analysis moment for each physiological modulation value; The mean and standard deviation of each physiological modulation value are calculated for all times within the preset analysis period. The sum of the mean and K times the standard deviation is used as the judgment threshold for each physiological modulation value; where K is a preset constant. For each physiological modulation value, the analysis time when the physiological modulation value is greater than the judgment threshold is selected as the effective peak time; the effective peak times corresponding to the EEG respiratory modulation value and the pulse respiratory modulation value are recorded as the EEG peak time and the pulse peak time, respectively.

9. The AI-based ERCP sedation complication prediction system according to claim 2, characterized in that, Linear fitting was performed on all data pairs corresponding to the period of stable body position using the least squares method.

10. The AI-based ERCP sedation complication prediction system according to claim 8, characterized in that, K is set to 1.

Images