Anesthesia pain prediction method and system based on virtual reality

Abstract

The invention discloses an anesthesia pain prediction method and system based on virtual reality, particularly relates to the technical field of medical virtual reality, and is used for solving the problems that in the prior art, in an anesthesia state, due to drug interference, a physiological signal background is changed, and the pain prediction accuracy is remarkably reduced. A predefined virtual reality stimulation sequence is applied to a patient through a virtual reality device, physiological signal data of the patient are synchronously collected, the change directions of various physiological signals before and after the virtual reality stimulation sequence are analyzed, and signal components conforming to a predefined collaborative response mode are screened to serve as candidate signal components; extracting a specific nerve physiological marker, and analyzing the time sequence relation of the incubation period and the dynamic change track consistency of various physiological signals; and when the electroencephalogram response latency is earlier than the facial myoelectricity activation latency and the dynamic change track consistency meets a preset condition, judging a high-specificity pain nerve marker, and calculating and outputting a prediction score of the patient on surgical pain.

Description

Technical Field

[0001] This invention relates to the field of medical virtual reality technology, and in particular to a method and system for predicting anesthesia and pain based on virtual reality. Background Technology

[0002] In the field of clinical anesthesia and pain management, existing technologies include methods for preoperative assessment using virtual reality devices. This typically involves exposing conscious patients to a standardized virtual reality environment, collecting and analyzing their physiological signals to predict their response to surgical pain and anesthesia requirements. Such methods are based on the premise that the patient's physiological response in the virtual reality environment can effectively map the sensitivity of their unaffected nervous system to potential pain stimuli.

[0003] However, when attempting to apply such predictive methods to the intraoperative stage, i.e., when the patient has received anesthetic drugs, the anesthetic drugs systematically alter the patient's central nervous system state, leading to significant changes in their physiological signal background. In this context, the specific neurophysiological markers related to pain perception induced by virtual reality stimulation have weak signal strength and are completely submerged in the strong background changes induced by drugs. Existing analytical methods that rely on establishing a physiological baseline or extracting signal features in a conscious state are unable to effectively separate the effective information for predicting pain response from the mixed signals heavily contaminated by drugs, resulting in a significant decrease in prediction accuracy. Summary of the Invention

[0004] This invention addresses the technical problems existing in the prior art by providing a method and system for predicting anesthesia pain based on virtual reality.

[0005] The technical solution of the present invention to solve the above-mentioned technical problems is as follows: Virtual reality-based methods for predicting anesthesia pain include: S1. While the patient is under anesthesia, a predefined virtual reality stimulation sequence is applied to the patient through a virtual reality device, and the patient's physiological signal data is collected simultaneously. S2. Analyze the changes in the direction of various types of physiological signals before and after the virtual reality stimulus sequence in the physiological signal data, and screen out the signal components that conform to the predefined cooperative response pattern as candidate signal components from the physiological signal data. S3. Extract neurophysiological features related to pain perception from candidate signal components as specific neurophysiological markers; S4. Analyze the temporal sequence relationship between the EEG response latency and the facial electromyography activation latency corresponding to specific neurophysiological markers, and analyze the consistency of the dynamic change trajectory of specific neurophysiological markers with multiple physiological signals. S5. When the latency of the EEG response is earlier than the latency of the facial electromyography activation, and the consistency of the dynamic change trajectory meets the preset conditions, a highly specific pain neuromarker is determined. S6. Calculate and output the patient's predicted score for surgical pain based on highly specific pain neuromarkers.

[0006] Further, step S1 includes: A predefined sequence of virtual reality stimuli, including visual and auditory stimuli, is applied to the patient using a virtual reality device. Simultaneously, the patient's physiological signal data, including electroencephalogram (EEG) signals, facial electromyography (EMG) signals, and skin conductance response signals, are collected.

[0007] Furthermore, visual stimulation simulates medical operation scenarios, while auditory stimulation provides environmentally relevant sound effects.

[0008] Further, step S2 includes: For electroencephalogram (EEG) signals, facial electromyography (EMG) signals, and skin conductance response signals, the direction of change of signal features before and after the virtual reality stimulation sequence was extracted; The consistency of the changing directions of various physiological signals with predefined collaborative response patterns is compared. The predefined collaborative response pattern defines the expected correlation between the directions of change in EEG signals, facial electromyography signals, and skin conductance response signals under painful stimuli; Based on the consistency comparison results, signal components corresponding to signal time periods whose change direction conforms to the predefined cooperative response pattern are selected from physiological signal data as candidate signal components.

[0009] Further, step S3 includes: Extracting the EEG response latency of EEG signals and the facial electromyography activation latency of facial electromyography signals from candidate signal components; Electroencephalogram (EEG) response latency and facial electromyography (EMG) activation latency were used as specific neurophysiological markers.

[0010] Further, step S4 includes: Compare the temporal sequence of EEG response latency and facial electromyography activation latency to determine the temporal order. Based on the continuous changes of EEG signals, facial electromyography signals, and skin conductance response signals during the virtual reality stimulation sequence, the signal change trajectories at corresponding time points of the EEG response latency and facial electromyography activation latency were extracted. Statistical consistency was calculated among the trajectories of EEG signal changes, facial electromyography signal changes, and skin conductance response signal changes to assess the consistency of dynamic trajectory changes.

[0011] Further, step S5 includes: When the latency of the EEG response is earlier than the latency of the facial electromyography (EMG) activation, and the statistical consistency between the EEG signal change trajectory, the facial EMG signal change trajectory, and the skin conductance response signal change trajectory exceeds a preset correlation threshold, the corresponding specific neurophysiological marker is identified as a high-specificity pain neuromarker.

[0012] Further, step S6 includes: The time difference between the EEG response latency and the facial electromyography activation latency corresponding to highly specific pain neuromarkers, as well as the statistical consistency between the EEG signal change trajectory and the skin conductance response signal change trajectory, are used as input features. The predicted score of the patient's surgical pain is calculated by a pre-established pain prediction regression model and then output.

[0013] Furthermore, the pre-established pain prediction regression model is obtained by using the time difference between the EEG response latency and the facial electromyography activation latency corresponding to high-specificity pain neuromarkers of multiple patients in the historical dataset, as well as the degree of statistical consistency between the EEG signal change trajectory and the skin conductance response signal change trajectory, as input features, and the corresponding real pain score as the output label, and is trained through regression analysis.

[0014] On the other hand, the present invention provides an anesthesia pain prediction system based on virtual reality, comprising: An application acquisition module is used to apply a predefined virtual reality stimulation sequence to a patient while the patient is under anesthesia, and simultaneously acquire the patient's physiological signal data. The signal filtering module is used to analyze the changes in the direction of various types of physiological signals in physiological signal data before and after virtual reality stimulus sequences, and to filter out signal components that conform to predefined cooperative response patterns from the physiological signal data as candidate signal components. The label extraction module is used to extract neurophysiological features related to pain perception from candidate signal components as specific neurophysiological markers. The labeling analysis module is used to analyze the temporal sequence relationship between the EEG response latency and the facial electromyography activation latency corresponding to specific neurophysiological markers, and to analyze the consistency of the dynamic change trajectory of specific neurophysiological markers with various physiological signals. The condition determination module is used to determine highly specific pain neuromarkers when the latency of the EEG response is earlier than the latency of the facial electromyography activation and the consistency of the dynamic change trajectory meets the preset conditions. The scoring output module is used to calculate and output a patient's predicted score for surgical pain based on highly specific pain neuromarkers.

[0015] The beneficial effects of this invention are: 1. By introducing virtual reality stimulation sequences under anesthesia and constructing a multi-physiological signal collaborative analysis framework, the interference of anesthetic drugs on the physiological signal background is effectively overcome. When the patient has received anesthetic drugs, the predefined virtual reality stimulation sequences can stimulate neural pathways related to pain perception. By analyzing the changes in various physiological signals before and after stimulation and screening candidate signal components that conform to the collaborative response pattern, specific responses related to pain can be accurately captured from the complex signal background affected by drugs. This significantly improves the sensitivity of identifying effective pain signals under anesthesia, allowing subsequent feature extraction and analysis processes to be based on purer and more relevant signals.

[0016] 2. By analyzing the temporal relationship between the latency of EEG response and facial electromyography activation, as well as the consistency of the dynamic change trajectories of multiple signals, a criterion for determining highly specific pain neuromarkers was constructed. This not only ensured the strong correlation between the extracted markers and pain neuropathies, but also effectively eliminated non-specific physiological fluctuations caused by anesthetic drugs through the dual constraints of temporal and synergistic change characteristics. Finally, the pain prediction score calculated based on these highly specific markers can accurately reflect the patient's true tendency to respond to surgical pain under anesthesia, providing a reliable basis for the precise adjustment of the intraoperative anesthesia plan and achieving the technical effect of maintaining high prediction accuracy even in complex drug interference environments. Attached Figure Description

[0017] Figure 1 This is a flowchart of the virtual reality-based anesthesia pain prediction method of the present invention; Figure 2 This is a schematic diagram of the structure of the virtual reality-based anesthesia pain prediction system of the present invention. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0019] Example 1: Figure 1 The present invention provides a virtual reality-based method for predicting anesthesia pain, comprising: S1. While the patient is under anesthesia, a predefined virtual reality stimulation sequence is applied to the patient through a virtual reality device, and the patient's physiological signal data is collected simultaneously. S2. Analyze the changes in the direction of various types of physiological signals before and after the virtual reality stimulus sequence in the physiological signal data, and screen out the signal components that conform to the predefined cooperative response pattern as candidate signal components from the physiological signal data. S3. Extract neurophysiological features related to pain perception from candidate signal components as specific neurophysiological markers; S4. Analyze the temporal sequence relationship between the EEG response latency and the facial electromyography activation latency corresponding to specific neurophysiological markers, and analyze the consistency of the dynamic change trajectory of specific neurophysiological markers with multiple physiological signals. S5. When the latency of the EEG response is earlier than the latency of the facial electromyography activation, and the consistency of the dynamic change trajectory meets the preset conditions, a highly specific pain neuromarker is determined. S6. Calculate and output the patient's predicted score for surgical pain based on highly specific pain neuromarkers.

[0020] In step S1, a predefined virtual reality stimulation sequence is applied to the patient via a virtual reality device. This sequence includes visual and auditory stimuli. The visual stimuli simulate medical procedures, such as generating a three-dimensional dynamic scene containing surgical instrument operations using computer graphics rendering technology. The scene content may include continuous animations simulating syringe puncture, scalpel cutting, or suture needle insertion. The duration of each scene is set to a specific value, for example, between 5 and 30 seconds, based on clinical validation data. The entire sequence consists of multiple scenes combined in a preset order, with the total duration controlled within the range of 1 to 5 minutes. The predefined virtual reality stimulation sequence is determined through previous clinical trials, for example, by collecting physiological response data from healthy subjects under different visual stimuli, selecting scene content that can stably induce pain-related neurophysiological responses, and arranging them in order of increasing stimulus intensity. The virtual reality device uses a head-mounted display with a resolution of, for example, 1920×1080 pixels and a refresh rate of, for example, 60 Hz, to ensure the smoothness and realism of the visual display and avoid visual fatigue or signal interference caused by insufficient device performance. Auditory stimulation is provided through high-fidelity headphones, which deliver environment-related sound effects, such as recorded instrument operation sounds, equipment alarm sounds, or digitally synthesized physiological sound effects from a real medical environment. The sound intensity is controlled within, for example, the range of 60 to 70 decibels. All sound effect files are standardized during production to ensure that their frequency characteristics conform to the range of human auditory sensitivity, for example, the main energy is concentrated between 500 Hz and 4000 Hz. Synchronization between auditory and visual stimuli is achieved through hardware synchronization signals, such as sending digital trigger pulses to the audio playback device when the visual scene changes, to ensure that the audio-visual synchronization error is less than 10 milliseconds.

[0021] Simultaneous acquisition of the patient's physiological signal data, including electroencephalogram (EEG) signals, facial electromyography (EMG) signals, and skin conductance response (SCR) signals, was performed. EEG signals were acquired using a multi-channel EEG acquisition system. Electrode placement followed the internationally accepted EEG recording standard 10-20 system, for example, placing acquisition electrodes at the Fz point in the frontal lobe, the Cz point in the central lobe, and the Pz point in the parietal lobe. Reference electrodes were placed on both earlobes. Silver chloride conductive gel was used between the electrodes and the scalp to maintain a contact impedance below 5 kΩ. The signal sampling rate was set to 256 Hz, and the acquisition bandwidth was set to 0.5 Hz to 40 Hz. Baseline signal recording was performed before formal acquisition, for example, continuously collecting data for 30 seconds while the patient was quiet with eyes closed as an individualized baseline. Facial... Electromyography (EMG) signals were acquired using a surface EMG acquisition device. Electrodes were attached to the surface projection area of ​​the corrugator supercilii muscle group on the face, using a bipolar lead configuration. The electrode spacing was precisely maintained at 2 cm. The signal sampling rate was set to 1000 Hz, and the acquisition bandwidth was set to 10 Hz to 500 Hz. After electrode attachment, the patient was asked to complete a set of standard facial movements, such as frowning and relaxing, to verify signal quality. Skin conductance response signals were acquired using a skin conductance response acquisition module. Sensors were fixed to the fingertips of the patient's index and ring fingers, using a constant voltage measurement method. The sampling rate was set to 10 Hz, and the measurement range was set to 0 to 50 microsiemens. Before acquisition, the patient was allowed to acclimatize in a constant temperature and humidity environment for 5 minutes to stabilize the baseline skin conductance level.

[0022] All physiological signal acquisition devices and virtual reality devices are synchronized with a unified clock source, such as a synchronization controller based on a network time protocol, to ensure that the synchronization accuracy of all signal timestamps is better than 1 millisecond. During the acquisition process, signal quality indicators are monitored in real time, such as whether the peak-to-peak amplitude of the EEG signal is consistently within the normal range of 50 microvolts to 100 microvolts, whether the baseline fluctuation of the facial electromyography signal in the resting state is less than 5 microvolts, and whether the baseline value of the skin conductance response signal is maintained between 2 microsiemens and 20 microsiemens. When any signal indicator exceeds the preset range, a prompt is immediately made to check the device connection.

[0023] The acquired raw signals are transmitted to the data processing unit in real time for preprocessing. The EEG signals are sequentially filtered by a 50 Hz notch filter to eliminate power frequency interference, a 0.5 Hz high-pass filter to eliminate baseline drift, and a 40 Hz low-pass filter to suppress high-frequency noise. The facial electromyography (EMG) signals are filtered by a 60 Hz notch filter and band-pass filters from 10 Hz to 500 Hz. The skin conductance response signals are filtered by a 0.1 Hz high-pass filter to eliminate long-term drift. All filtering processes use Butterworth filters of the same order to ensure consistent phase response. The preprocessed signals are stored in floating-point format. Each data point is associated with a timestamp accurate to the millisecond level and a corresponding stimulus event marker to form a complete time-domain signal sequence for subsequent analysis steps.

[0024] In step S2, the direction of change of signal features before and after the virtual reality stimulation sequence is extracted for the electroencephalogram (EEG), facial electromyography (EMG), and electrodermal response (EDR) signals. Specifically, for the EEG signal, a time window before the start of the virtual reality stimulation sequence is defined as the baseline period, for example, from 5 seconds before the start of the stimulation to the start of the stimulation, and a time window after the end of the stimulation sequence is defined as the response period, for example, from 5 seconds after the end of the stimulation to 10 seconds after the end of the stimulation. The length of the baseline and response periods is set based on signal stability and clinical experience. For example, previous experiments have verified that a 5-second window can effectively capture physiological responses without being affected by random fluctuations. The choice of window length takes into account the signal sampling rate and the delayed characteristics of physiological responses. For example, the EEG signal response is relatively slow, so a 5-second window is used to smooth short-term fluctuations, while the facial EMG signal response is relatively fast, but the same window setting is used to maintain consistency. During the baseline period, the power spectral density of the EEG signal in a specific frequency band is calculated, such as the power values ​​in the theta band from 4 Hz to 8 Hz. The signal is converted from the time domain to the frequency domain by Fast Fourier Transform (FFT). The FFT window length is set to, for example, 2 seconds, with an overlap rate of 50% to balance frequency resolution and time resolution. When calculating the power spectral density, the signal is windowed, for example, using a Hanning window to reduce spectral leakage. Then, the arithmetic mean of the power values ​​of all windows during the baseline period is taken as the baseline value.

[0025] During the response period, the same frequency band and calculation method are used to obtain the average power spectral density of the response period as the response value. Then, the baseline value and the response value are compared. If the response value is greater than the baseline value, the direction of change is marked as increasing; if the response value is less than the baseline value, the direction of change is marked as decreasing. The extraction of the direction of change is based on simple numerical comparison, which is implemented through programming to make conditional judgments, such as using greater than or less than operators in the software for comparison. For facial electromyography signals, the same baseline and response period definitions are used, and the root mean square value of the signal is calculated as a feature. The root mean square value is obtained by squaring the signal value, taking the arithmetic mean, and then taking the square root of the mean. The average root mean square values ​​of the baseline and response periods are compared. If the average value of the response period is greater than the average value of the baseline period, the direction of change is marked as increasing; otherwise, it is marked as decreasing. The calculation of the root mean square value considers the entire signal band and does not perform frequency band filtering to capture the overall changes in muscle activity.

[0026] For the skin conductance response signal, the basal and response phases are set identically. The average skin conductance level is calculated directly from the preprocessed value of the acquired raw signal. Preprocessing includes removing baseline drift and motion artifacts, for example, using a high-pass filter with a cutoff frequency of 0.1 Hz. The average values ​​of the basal and response phases are compared to determine whether the change direction is increasing or decreasing. The direction of change extraction of the skin conductance response signal is based on the DC component, ignoring high-frequency fluctuations. The extraction process for the direction of change of all signal features is time-aligned, using the same timestamp to ensure consistency. The timestamp comes from the synchronization signal of the virtual reality device. The results of the direction of change are stored as discrete values ​​increasing or decreasing for subsequent steps. The storage format is a list or array, with each element corresponding to a signal type and time period.

[0027] A predefined co-response model defines the expected association between the directions of change in EEG, facial electromyography (EMG), and electrodermal response (EDR) signals under painful stimuli. This model is established by analyzing historical clinical data, such as collecting physiological signal records and corresponding pain scores from multiple patients under standard painful stimuli. Pain scores are derived from patient self-reports or physician assessments, using a numerical scale such as 0 to 10, where 0 represents no pain and 10 represents severe pain. Historical data includes samples from different populations, such as those aged 20 to 60 years, with a balanced male-to-female ratio to ensure the model's generalization. Within the historical data, the co-occurrence frequency of changes in the directions of change in EEG, EMG, and ERD signals during painful events is statistically analyzed. A painful event is defined as a pain score exceeding a threshold, such as 7. The co-occurrence frequency is calculated as the conditional probability that all three signal change directions simultaneously increase. For example, this is achieved by counting the number of times the signal directions are consistent during a painful event and dividing by the total number of painful events. If the probability exceeds a preset value, such as 0.8, the expected association is set as an increase in the direction of change in EEG, facial EMG, and ERD signals. A predefined collaborative response pattern can be represented as a triple, such as increase, increase, increase, corresponding to EEG, facial electromyography, and skin conductance response, respectively. Pattern establishment is based on a data-driven approach, such as using a decision tree algorithm to analyze the association rules between signal change direction and pain events in historical data. Parameters of the decision tree algorithm, such as the minimum number of sample splits, are set to 10 and the maximum depth to 5 to prevent overfitting. The rule with the highest support and confidence is selected as the pattern. Support is defined as the frequency of a rule's occurrence in the dataset, and confidence is defined as the probability that the rule is correct. The pattern content is embedded in the analysis system before implementation to ensure that the same expected association is used in each comparison. Embedding is achieved by storing the pattern as a configuration file or database record, which is loaded when the system starts.

[0028] The consistency of various physiological signal change directions is compared with predefined coordinated response patterns. The comparison process is achieved by calculating the matching degree between the actual extracted change directions and the predefined patterns. For example, for each signal time period, the actual change directions form a vector (e.g., EEG change direction, facial electromyography change direction, skin conductance response change direction), and the predefined pattern forms another vector. The matching degree is calculated by dividing the number of consistent elements in the two vectors by the total number of elements (3). The number of consistent elements is the number of times the actual change direction matches the predefined pattern, and the matching degree value ranges from 0 to 1. Consistency comparison can also employ a weighted approach, such as assigning weights based on the importance of the signal. The weight values ​​are set based on the contribution of each signal to pain prediction in historical data. The contribution is obtained through regression analysis, such as using a linear regression model to calculate the correlation coefficient between each signal change direction and the pain score, and then normalizing the correlation coefficient to a weight. However, in this embodiment, a simple average is used for simplification, meaning each signal has an equal weight. The comparison output is a consistency score, representing the degree of conformity between the actual signal change and the expected pattern. Consistency alignment algorithms are implemented programmatically, for example, by using conditional statements to compare the direction of change of each signal, counting consistent occurrences, and then calculating the degree of match. The alignment process considers the continuity of signal time periods, for example, by averaging overlapping windows to reduce noise.

[0029] Based on the consistency comparison results, signal components corresponding to signal periods whose change direction conforms to a predefined cooperative response pattern are selected from physiological signal data as candidate signal components. Specifically, a consistency threshold is set, for example, 0.7. The threshold is determined through prior validation data, such as using receiver operating characteristic (ROC) curve analysis, to select a threshold that balances sensitivity and specificity on independent datasets. The target values ​​for sensitivity and specificity are set to, for example, above 0.8. Threshold adjustment is based on cross-validation methods, such as dividing the data into training and test sets, calculating performance indicators under different thresholds on the training set, and selecting the threshold that performs best on the test set. When the consistency comparison result is greater than or equal to the threshold, the change direction of the signal period is considered to conform to the predefined cooperative response pattern. A signal period refers to a continuous time interval during a virtual reality stimulus sequence. For example, the entire sequence is divided into non-overlapping windows, each with a length of 10 seconds. The change direction is extracted and consistency is compared independently for each window. The screening process traverses all windows, selecting windows with a consistency score reaching the threshold, and extracting the corresponding raw signal data or preprocessed signal components as candidate signal components. Candidate signal components include EEG signal segments, facial electromyography (EMG) signal segments, and skin conductance response (SCR) signal segments, which are time-aligned and accompanied by consistency score information. The selection logic is implemented through an iterative loop; for example, a consistency score is calculated for each window and compared with a threshold. If the condition is met, the window is marked as a candidate. The threshold and window parameters can be adjusted according to the application scenario. For example, a shorter window length, such as 5 seconds, may be used in pediatric patients to accommodate faster physiological changes. Adjustments are based on clinical expert opinions or previous trial data. The selected candidate signal components are used for further processing in subsequent steps and stored as time-series data blocks. Each block contains signal values ​​and metadata such as timestamps and consistency scores.

[0030] In step S3, the EEG response latency of the EEG signal is extracted from the candidate signal components. The candidate signal components are EEG signal segments corresponding to the signal time periods selected in step S2. The EEG signal segments are preprocessed continuous time series data with timestamps. Preprocessing includes filtering and denoising to ensure signal quality. The EEG response latency is defined as the time interval from the start of the virtual reality stimulus sequence to the appearance of a significant response in the EEG signal. A significant response refers to the moment when the signal feature value exceeds a preset threshold. The threshold is set based on the statistical characteristics of the signal during the baseline period. The extraction process first performs band filtering on the EEG signals in the candidate signal components. For example, a digital bandpass filter is used to retain components in the theta band from 4 Hz to 8 Hz to focus on neural activity related to pain perception. The filter design is based on the infinite impulse response structure, and the cutoff frequency setting is determined through previous experiments, such as testing the sensitivity of different frequency bands to pain response on independent datasets and selecting the frequency band with the most significant power change in pain events. Then, the power spectral density of the filtered signal is calculated using a short-time Fourier transform (SFT) method. The window length is set to, for example, 2 seconds, and the overlap rate is set to 50% to balance time and frequency resolution. The power spectral density is calculated by segmenting the signal, applying a window (e.g., a Hanning window), performing a Fourier transform, and taking the square of the modulus. The use of the Hanning window reduces spectral leakage and improves the accuracy of frequency estimation. The baseline period is defined as a time window before the start of the virtual reality stimulus sequence, for example, from 5 seconds before the stimulus begins to the start of the stimulus. The mean and standard deviation of the theta band power spectral density within the baseline period are calculated as reference statistics. The mean and standard deviation are calculated using arithmetic methods to ensure that the statistics reflect the baseline level of the signal in the unstimulated state. The response threshold is set to the mean of the baseline period plus twice the standard deviation. This multiple is determined based on the analysis of signal noise levels in historical data, for example, by calculating the standard deviation distribution of power fluctuations in healthy subjects in the unstimulated state, and selecting a multiple that covers 95% of the cases to minimize the risk of false detection.

[0031] Starting from the beginning of the virtual reality stimulus sequence, the power spectral density value at each time point is checked sequentially. When the power value first consistently exceeds the response threshold for a preset duration, such as 100 milliseconds, that time point is marked as the response start time. The duration is set to avoid false detections caused by short fluctuations. This is achieved through a sliding window verification, for example, using a 100-millisecond window to check for consecutive exceedances of the threshold. The EEG response latency is calculated as the difference between the response start time and the stimulus start time, in milliseconds, to ensure accuracy conforms to the time scale of neural responses. The calculation result is stored as a numerical variable with an identifier for the corresponding signal period. If no significant response is detected, for example, if the power value does not consistently exceed the threshold within 5 seconds after the stimulus begins within the preset time range, that period is marked as no response, and the latency is recorded as a null value or a specific identifier to avoid misuse in subsequent steps.

[0032] The facial electromyography (EMG) activation latency is extracted from candidate signal components. Candidate signal components include the facial EMG signal segments selected in step S2. These segments are preprocessed time-series data, and the preprocessing includes bandpass filtering and motion artifact removal. The facial EMG activation latency is defined as the time interval from the start of the virtual reality stimulus sequence to the appearance of significant activation in the facial EMG signal. Significant activation refers to the moment when the signal feature value exceeds a preset threshold. The extraction process first performs full-band processing on the facial EMG signals in the candidate signal components without frequency band filtering to preserve the complete information of muscle activity. However, a bandpass filter, such as 10 Hz to 500 Hz, is applied to remove low-frequency drift and high-frequency noise. The filter parameters are set based on the physiological characteristics of the electromyography signal.

[0033] The root mean square (RMS) value of the signal is calculated as a feature. The RMS value is obtained by squaring the signal value, taking the arithmetic mean, and then taking the square root of the mean. The calculation window length is set to, for example, 100 milliseconds, and the sliding step size is set to 50 milliseconds to capture rapid changes in muscle activity. The selection of the window and step size is based on previous experimental data, such as testing the signal's response speed to pain activation under different window conditions. The basal period is defined as a time window before the start of the virtual reality stimulus sequence, for example, from 5 seconds before the stimulus begins to the moment of stimulus onset. The mean and standard deviation of the RMS value within the basal period are calculated as reference statistics. The mean and standard deviation are calculated using the same window parameters to ensure consistency. The activation threshold is set to the basal period mean plus 3 times the standard deviation. This multiple is determined based on the activity fluctuation analysis of facial muscles in a resting state, for example, adjusted by the coefficient of variation of the signal in historical data without pain stimulation, selecting a multiple with a false alarm rate of less than 5% on an independent validation set.

[0034] Starting from the beginning of the virtual reality stimulus sequence, the root mean square (RMS) value at each time point is checked sequentially. When the RMS value first consistently exceeds the activation threshold for a preset duration (e.g., 50 milliseconds), that time point is marked as the activation start time. The duration is set considering the physiological characteristics of muscle response and is selected through experimental verification, such as observing the typical duration of muscle activation in clinical data. The facial electromyography (EMG) activation latency is calculated as the difference between the activation start time and the stimulus start time, in milliseconds. The calculation result is stored as a numerical variable with an identifier for the corresponding signal period. During extraction, if multiple candidate signal components exist, the latency is calculated independently for each component, and all results are recorded for subsequent analysis. If no activation is detected, it is marked as no activation, and the latency is set to a null value.

[0035] EEG response latency and facial electromyography (EMG) activation latency are used as specific neurophysiological markers, which are specific physiological indicators used to characterize pain perception. After being extracted from candidate signal components, EEG response latency and EMG activation latency are directly stored as marker values ​​in a numeric pair format, such as two fields in a data structure corresponding to EEG response latency and EMG activation latency, respectively. The markers are defined based on their association with pain neural pathways; EEG response latency reflects the early processing speed of the central nervous system, while EMG activation latency reflects the reflexive response of peripheral muscles. The combination of these two provides complementary information. The use of markers ensures compatibility with subsequent steps, such as in step S4 for consistency analysis of temporal relationships and dynamic change trajectories. All marker data is accompanied by metadata, such as identifiers of the source signal time period, parameter settings used during extraction, and timestamp information, to ensure traceability and consistency. The assignment of marker values ​​is implemented programmatically, for example, binding latency values ​​to specific variable names in software and outputting them as structured data formats such as JSON or arrays for easy retrieval in subsequent steps. If any latency period is null, the marker is considered invalid and will not be used in subsequent analysis to avoid introducing noisy data.

[0036] In step S4, the temporal order is determined by comparing the EEG response latency and the facial electromyography (EMG) activation latency. The EEG response latency and the EMG activation latency are derived from specific neurophysiological markers extracted in step S3. These specific neurophysiological markers are numerical data, all in milliseconds, representing the time interval from the start of the virtual reality stimulus sequence to the appearance of a significant response in the EEG signal or significant activation in the facial EMG signal. The temporal order is determined through direct numerical comparison. The difference between the EEG response latency and the EMG activation latency is calculated. A negative difference indicates that the EEG response latency is earlier than the EMG activation latency, while a positive difference indicates that the EEG response latency is later than the EMG activation latency. The difference is calculated using arithmetic subtraction, and the result is stored as a signed numerical variable.

[0037] The definition of temporal sequence relationships is based on the principles of pain neurophysiology. For example, in the process of pain perception, the electrophysiological response of the central nervous system usually occurs before the reflexive activation of peripheral muscles. Therefore, the "earlier than" relationship specifically refers to the situation where the latency value of the EEG response is less than the latency value of the facial electromyography activation. The comparison process considers the accuracy range of the latency measurement values. For example, due to the sampling rate limitations of the signal acquisition equipment and the time alignment error in the preprocessing process, the latency value may have measurement uncertainty at the millisecond level. Therefore, when the absolute value of the difference is less than a preset tolerance threshold, such as 5 milliseconds, the temporal sequence relationship can be marked as approximately simultaneous. The tolerance threshold is set based on the temporal resolution characteristics of the signal system and is determined through the analysis of previous experimental data. For example, the smallest resolvable unit of latency difference in multiple measurements is statistically analyzed.

[0038] Based on the continuous changes in EEG signals, facial electromyography (EMG) signals, and electrodermal response (EDR) signals during virtual reality stimulation sequences, signal change trajectories are extracted at corresponding time points of the EEG response latency and facial EMG activation latency. The EEG, EMG, and EMR signals are derived from the physiological signal data acquired in step S1, which is preprocessed and stored as a timestamped continuous time series. The signal change trajectory is defined as the continuous change sequence of each signal characteristic value within a time window centered on the latency time point. For example, for the EEG response latency time point, the time window is selected as the interval from 50 milliseconds before to 100 milliseconds after that time point. The window length is determined based on the typical duration characteristics of the physiological response signal, by analyzing the waveform duration distribution of pain-related neurophysiological responses in historical data.

[0039] The extraction process first establishes a unified temporal reference system, using the start time of the virtual reality stimulus sequence as the time zero point. Timestamps of all signals are aligned and corrected to ensure synchronization of the timelines for different signal types. For EEG signals, specific frequency band power spectral density values ​​are used as feature values, such as power values ​​in the theta band within the range of 4 Hz to 8 Hz. The power spectral density is calculated using a short-time Fourier transform method, with a transform window length set to, for example, 2 seconds and an overlap rate set to 50%, to maintain continuity in the temporal dimension. For facial electromyography (EMG) signals, root mean square (RMS) values ​​are used as feature values, with a calculation window length set to, for example, 100 milliseconds and a sliding step size set to 50 milliseconds, to preserve the rapid changes in muscle electrical activity.

[0040] For the skin conductance response signal, the feature values ​​are either the original measured values ​​of skin conductance or smoothed values. The smoothing is achieved using a moving average method, with the moving average window length set to, for example, 1 second. At the time point corresponding to the EEG response latency, the sequence of EEG signal feature values ​​within a defined window before and after that time point is extracted as the EEG signal change trajectory. Similarly, at the time point corresponding to the facial electromyography (EMG) activation latency, the sequence of facial EMG signal feature values ​​within the corresponding window is extracted as the facial EMG signal change trajectory. The extraction time window for the skin conductance response signal change trajectory starts from the earlier of the EEG response latency and the facial EMG activation latency to ensure that the trajectory data covers the complete physiological response period.

[0041] The extracted trajectory data is stored as a numerical array structure, with each array element corresponding to a feature value at a specific time point. The original temporal resolution is consistent with the sampling rate of each signal acquisition; for example, 256 Hz for EEG signals, 1000 Hz for facial electromyography (EMG) signals, and 10 Hz for skin conductance. To enable direct comparison between different signal trajectories, all trajectory data are resampled to a uniform temporal frequency, such as 100 Hz, using linear interpolation. After trajectory data extraction, corresponding timestamp information and source latency values ​​are appended to form a complete data structure for subsequent analysis.

[0042] Statistical consistency among EEG signal change trajectories, facial electromyography (EMG) signal change trajectories, and electrodermal response (EDR) signal change trajectories is calculated to assess the consistency of dynamic trajectory changes. Statistical consistency refers to the degree of similarity in the morphological change trends of multiple signal change trajectories. The assessment method is achieved by calculating the correlation coefficient between each trajectory sequence, for example, using the Pearson product-moment correlation coefficient to quantify the degree of linear correlation. The input parameters are the three signal change trajectory data extracted in the preceding steps: numerical sequences of EEG signal change trajectories, facial EMG signal change trajectories, and ERD signal change trajectories. These numerical sequences have the same temporal length and uniform temporal resolution, and time axis alignment ensures data comparability.

[0043] The calculation process first performs normalization preprocessing on the trajectory data. The numerical value of each trajectory sequence is subtracted from its arithmetic mean and then divided by the sequence's standard deviation to eliminate differences in the dimensions and amplitude ranges of different signal features. Then, the Pearson correlation coefficients between every two trajectory sequences are calculated, including the correlation coefficients between EEG signal changes and facial electromyography (EMG) signal changes, EEG signal changes and skin conductance response (SCR) signal changes, and facial EMG signal changes and SCR signal changes.

[0044] The Pearson correlation coefficient is calculated by dividing the product of the covariance of the two sequences by their respective standard deviations. The covariance is calculated using the standard arithmetic formula: the sum of the products of the differences between corresponding elements of the two sequences divided by the sequence length minus 1. The evaluation result of the consistency of the dynamic trajectory is represented by a comprehensive consistency score, which is obtained by calculating the arithmetic mean of the three correlation coefficients. The mean is calculated using a simple arithmetic mean method, with each correlation coefficient having equal weight.

[0045] The threshold for the overall consistency score is set based on prior validation data analysis. For example, the receiver operating characteristic (ROC) curve method is used to analyze the sensitivity and specificity of different thresholds for pain event recognition on independent datasets, selecting the optimal threshold that balances performance. The threshold value typically ranges from 0.5 to 0.8. If trajectory data is missing or invalid during the evaluation process, such as incomplete trajectory sequences due to signal acquisition interruptions, the statistical consistency calculation for that time period is skipped, and that period is marked as invalid data to avoid erroneous consistency evaluation results. The entire statistical consistency calculation process is implemented through a programmed loop structure, processing the trajectory data for each valid time period sequentially and calling mathematical calculation functions to solve for the correlation coefficient.

[0046] In step S5, when the EEG response latency is earlier than the facial electromyography (EMG) activation latency, the first judgment condition check procedure is executed. The EEG response latency and the facial EMG activation latency are derived from specific neurophysiological markers extracted in step S3. These specific neurophysiological markers are numerical data, all in milliseconds, representing the time interval from the start of the virtual reality stimulus sequence to the appearance of a significant response in the EEG signal or a significant activation in the facial EMG signal. The earlier-than-expected relationship is determined by comparing the magnitudes of the EEG response latency and the facial EMG activation latency. Specifically, the arithmetic difference between the EEG response latency and the facial EMG activation latency is calculated. If the difference is less than zero, the EEG response latency is determined to be earlier than the facial EMG activation latency. The difference is calculated using standard arithmetic subtraction, and the result is stored as a signed numerical variable for subsequent logical judgment. The determination process needs to consider the accuracy characteristics of the measurement system. For example, due to differences in sampling rates between EEG and facial electromyography (EMG) acquisition devices and time alignment errors in the preprocessing stage, the latency value may have a certain range of measurement uncertainty. Therefore, when the absolute value of the difference is less than a preset tolerance threshold, it is still considered to meet the earlier condition. The preset tolerance threshold is determined through analysis of previous experimental data. For example, by statistically analyzing the standard deviation distribution of latency differences in multiple repeated measurements, the tolerance threshold is set to twice the standard deviation, for example, 5 milliseconds, to ensure that the sequential relationship can still be correctly identified within the measurement error range. The physiological basis of the earlier relationship lies in the neural conduction characteristics of pain perception, that is, the electrophysiological response of the central nervous system usually occurs before the reflexive activation of peripheral muscles. This temporal characteristic is used to screen pain response patterns that conform to neurophysiological laws.

[0047] When the statistical consistency among the EEG signal change trajectory, facial electromyography (EMG) signal change trajectory, and skin conductance response (SCRR) signal change trajectory exceeds a preset correlation threshold, the second judgment condition check procedure is executed. The statistical consistency is derived from the comprehensive consistency score calculated in step S4. This score is obtained by taking the arithmetic mean of the Pearson correlation coefficients among the three trajectories, with a value ranging from -1 to +1. The preset correlation threshold is determined through validation using historical clinical data. Receiver operating characteristic (ROC) curve analysis is used to analyze the performance of different threshold values ​​for pain event recognition on independent datasets. The optimal threshold value that balances sensitivity and specificity is selected. Sensitivity reflects the ability to correctly identify real pain events, while specificity reflects the ability to correctly exclude non-pain events. The threshold value is typically set between 0.5 and 0.8, with specific values ​​such as 0.6 adjusted based on signal quality and clinical needs in the actual application scenario. The setting of the correlation threshold also needs to consider the influence of the signal acquisition environment. For example, in an operating environment with strong electromagnetic interference, the threshold requirement may be appropriately reduced, but prior testing is required to verify the applicability and stability of the threshold under different environmental conditions. The determination of whether statistical consistency exceeds the threshold is achieved by directly comparing the overall consistency score with the preset correlation threshold. If the overall consistency score is greater than or equal to the preset correlation threshold, it is considered to meet the condition. The comparison operation uses the standard numerical size judgment method, and the result is converted into a Boolean logic value for subsequent condition combination evaluation.

[0048] When both of the above conditions are met simultaneously—that is, when the EEG response latency is earlier than the facial electromyography (EMG) activation latency and the statistical consistency exceeds a preset correlation threshold—the corresponding specific neurophysiological marker is identified as a high-specificity pain neurophysiological marker. Specific neurophysiological markers refer to the EEG response latency and facial EMG activation latency value pairs extracted in step S3. The determination process is achieved by combining the results of the two conditions through logical AND operations. For example, in program design, a conditional statement is used to check if the "earlier than" relation is true and the consistency exceeding the threshold is also true before executing the determination operation.

[0049] The definition of highly specific pain neuromarkers is based on their strong correlation with pain neural pathways. These markers need to simultaneously reflect the temporal characteristics of central nervous system responses and the synergistic characteristics of multiple physiological signal changes, thereby improving the specificity of pain prediction. The determination result marks the current specific neurophysiological marker as a highly specific category, for example, by adding a specific state identifier to the data structure or outputting it to a dedicated storage area. The marking information includes metadata such as complete latency values, consistency scores, and determination timestamps. If either of the two conditions is not met, the original state of the specific neurophysiological marker is maintained, and no category upgrade is performed to avoid introducing the risk of misjudgment. The entire determination process needs to ensure the complete traceability of the data, such as recording the specific threshold parameters used in the determination and the time nodes of the condition checks, to facilitate subsequent verification analysis and quality control. Highly specific pain neuromarkers will be used as key input parameters for the prediction score calculation in step S6, completing the closed-loop processing of the pain assessment process.

[0050] In step S6, the time difference between the EEG response latency and the facial electromyography (EMG) activation latency corresponding to the high-specificity pain neurotransmitter is used as one of the input features. The high-specificity pain neurotransmitter comes from the judgment result of step S5, which includes the specific values ​​of the EEG response latency and the facial EMG activation latency. The specific values ​​are the specific neurophysiological markers extracted from the candidate signal components in step S3, and the unit is milliseconds. The time difference is calculated by subtracting the facial EMG activation latency value from the EEG response latency value. The calculation process uses standard arithmetic subtraction, and the result is stored as a signed numerical variable, with the unit remaining in milliseconds, to accurately reflect the time interval difference between the two latencies. The positive or negative sign of the time difference indicates the sequential relationship between the EEG response latency and the facial EMG activation latency. For example, a negative value indicates that the EEG response latency is earlier than the facial EMG activation latency, and a positive value indicates that it is later. However, in this step, the difference value is directly used as the feature input without additional direction judgment; only the magnitude of the value is considered for model prediction. The extraction of time difference is based on the incubation period data defined in step S3, ensuring that the data source is consistent and there is no need to recalculate. It can be obtained by simply performing numerical operations. The operation process is implemented through programming, for example, by using variables to store the incubation period values ​​in the software and performing subtraction operations, and the result is saved in floating-point format.

[0051] Simultaneously, the statistical consistency between the EEG signal change trajectory and the skin conductance response signal change trajectory is used as another input feature. The statistical consistency score is derived from the comprehensive consistency score calculated in step S4. This score is obtained by taking the arithmetic mean of the Pearson correlation coefficients between the EEG signal change trajectory, the facial electromyography (EMG) signal change trajectory, and the skin conductance response signal change trajectory, with a value ranging from -1 to +1. The statistical consistency score is obtained directly from the output of step S4 without additional processing, but it is crucial to ensure that the data format matches the model input requirements; for example, the score is converted to standard floating-point form and stored in a designated variable. The combination of input features forms a feature vector, including the time difference value and the statistical consistency score. These two features jointly characterize the temporal properties and multi-signal synergy of the pain response, providing a basis for prediction. The feature vector is constructed through data concatenation; for example, the time difference value and the statistical consistency score are stored sequentially in an array or list structure, ensuring the vector elements are in the same order for easy model processing.

[0052] The pre-established pain prediction regression model was trained using the time difference between the EEG response latency and facial electromyography activation latency corresponding to highly specific pain neuromarkers from multiple patients in a historical dataset, as well as the statistical consistency between the EEG signal change trajectory and the skin conductance response signal change trajectory. The corresponding actual pain scores were used as output labels. The historical dataset was constructed by collecting clinical experimental data, such as recording physiological signals from multiple patients under standard pain stimulation conditions and simultaneously acquiring actual pain scores. The number of patient samples was determined based on statistical power analysis, for example, including more than 100 patients to ensure data representativeness. The patient group encompassed different ages, genders, and pain sensitivities to enhance the model's generalization ability. Actual pain scores were obtained from patient self-reports or clinician assessments, using a numerical scale such as 0 to 10, where 0 represents no pain and 10 represents severe pain. Scoring was performed immediately after the stimulation to reduce memory bias. Data preprocessing includes cleaning outliers, such as removing records with time differences or statistical consistency exceeding a reasonable range. The reasonable range is determined by historical data distribution; for example, a time difference absolute value greater than 500 milliseconds or a statistical consistency less than -0.5 is considered an anomaly. Regression analysis uses the least squares method for model fitting. The least squares method solves for model parameters by minimizing the sum of squared errors between predicted and true values ​​through an iterative optimization process. The model form is a linear regression equation, for example, predicted score = parameter 1 × time difference + parameter 2 × statistical consistency + constant term. Parameter values ​​are obtained through training with historical data. During training, input features are standardized, for example, by subtracting the mean from the time difference and statistical consistency values ​​and then dividing by the standard deviation to eliminate the influence of dimensions. Standardized parameters are calculated based on training data and saved for subsequent predictions. After model training is complete, the parameters are fixed in the system, for example, stored as a coefficient matrix and intercept term, ensuring direct access during prediction. The model's performance is evaluated through cross-validation, for example, by randomly dividing historical data into training and test sets. The training set is used for parameter estimation, and the test set is used to calculate the correlation coefficient or root mean square error between the predicted score and the actual score to verify the model's accuracy and generalization ability. Performance metrics such as a correlation coefficient greater than 0.7 are considered acceptable.

[0053] A pre-established pain prediction regression model calculates a predicted pain score for the patient during surgery. The calculation process inputs the patient's current time difference and statistical consistency score into the model. The model performs linear combination operations based on stored parameters, such as multiplying the time difference by a corresponding coefficient, adding the statistical consistency score by a corresponding coefficient, and adding a constant term to obtain the predicted score. The output predicted score ranges from 0 to 10, representing the predicted pain intensity, consistent with the actual pain score. The output is presented via a display device or storage system, such as displaying the value on a user interface or saving it to a file for clinical reference. The entire calculation process ensures real-time performance, completing the prediction within milliseconds of receiving the input features, and includes a timestamp and patient identifier to maintain data traceability. If input features are missing or invalid, such as outliers in the time difference or statistical consistency score, an error message or default value is output to avoid erroneous predictions. The predicted score completes the pain assessment process, providing a basis for subsequent medical decisions.

[0054] Example 2: Figure 2 A schematic diagram of the structure of the virtual reality-based anesthesia pain prediction system of the present invention is given. The virtual reality-based anesthesia pain prediction system includes: An application acquisition module is used to apply a predefined virtual reality stimulation sequence to a patient while the patient is under anesthesia, and simultaneously acquire the patient's physiological signal data. The signal filtering module is used to analyze the changes in the direction of various types of physiological signals in physiological signal data before and after virtual reality stimulus sequences, and to filter out signal components that conform to predefined cooperative response patterns from the physiological signal data as candidate signal components. The label extraction module is used to extract neurophysiological features related to pain perception from candidate signal components as specific neurophysiological markers. The labeling analysis module is used to analyze the temporal sequence relationship between the EEG response latency and the facial electromyography activation latency corresponding to specific neurophysiological markers, and to analyze the consistency of the dynamic change trajectory of specific neurophysiological markers with various physiological signals. The condition determination module is used to determine highly specific pain neuromarkers when the latency of the EEG response is earlier than the latency of the facial electromyography activation and the consistency of the dynamic change trajectory meets the preset conditions. The scoring output module is used to calculate and output a patient's predicted score for surgical pain based on highly specific pain neuromarkers.

[0055] All calculations involved in the embodiments are dimensionless numerical calculations, and the preset parameters and thresholds in the calculations are set by those skilled in the art according to the actual situation.

[0056] It should be noted that this invention can be deployed on the device itself to realize embedded applications, or it can run on a PC or other terminal with a user interface, thereby meeting various hardware environments and usage requirements.

[0057] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions according to the embodiments of this application are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wireless or wired transmission; wired transmission methods include optical fiber, twisted pair, coaxial cable, etc.; wireless transmission includes infrared, microwave, etc. Computer-readable storage media can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more sets of available media. Available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media. Semiconductor media can be solid-state drives.

[0058] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0059] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or modules may be electrical, mechanical, or other forms.

[0060] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical modules; they may be located in one place or distributed across multiple network modules. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0061] In addition, the functional modules in the various embodiments of this application can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0062] If a function is implemented as a software module and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0063] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0064] In conclusion, the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A virtual reality-based anesthetic pain prediction method, characterized by, The method comprises the following steps: S1, in the case where the patient receives anesthetic drugs, a pre-defined virtual reality stimulation sequence is applied to the patient through a virtual reality device, and physiological signal data of the patient is synchronously collected; S2, the change directions of multiple types of physiological signals in the physiological signal data before and after the virtual reality stimulation sequence are analyzed, and signal components corresponding to signal periods in which the change directions conform to a pre-defined coordinated response mode are screened out from the physiological signal data as candidate signal components; S3, neural electrophysiological features related to pain perception are extracted from the candidate signal components as specific neurophysiological markers; S4, the time sequence relationship between the brain electrical response latency corresponding to the specific neurophysiological markers and the facial electromyographic activation latency is analyzed, and the dynamic change trajectory consistency of the specific neurophysiological markers and multiple physiological signals is analyzed; S5, when the brain electrical response latency is earlier than the facial electromyographic activation latency, and the dynamic change trajectory consistency meets a pre-set condition, a high-specificity pain neural marker is determined; S6, a prediction score of the patient's pain during the operation is calculated based on the high-specificity pain neural marker and output.

2. The virtual reality-based anesthetic pain prediction method of claim 1, wherein, Step S1 comprises: A pre-defined virtual reality stimulation sequence containing visual stimulation and auditory stimulation is applied to the patient through a virtual reality device; And physiological signal data of the patient, including electroencephalogram signals, facial electromyogram signals and galvanic skin response signals, is synchronously collected.

3. The virtual reality-based anesthetic pain prediction method of claim 2, wherein, The visual stimulation simulates a medical operation scene, and the auditory stimulation provides sound effects related to the environment.

4. The virtual reality-based anesthetic pain prediction method of claim 2, wherein, Step S2 comprises: The change directions of the signal features before and after the virtual reality stimulation sequence are extracted for the electroencephalogram signals, facial electromyogram signals and galvanic skin response signals, respectively; The change directions of the multiple physiological signals are compared with a pre-defined coordinated response mode for consistency; The pre-defined coordinated response mode defines the expected correlation between the change directions of the electroencephalogram signals, facial electromyogram signals and galvanic skin response signals under pain stimulation; Based on the consistency comparison result, signal components corresponding to signal periods in which the change directions conform to the pre-defined coordinated response mode are screened out from the physiological signal data as candidate signal components.

5. The virtual reality-based anesthetic pain prediction method of claim 4, wherein, Step S3 comprises: The brain electrical response latency of the electroencephalogram signals and the facial electromyographic activation latency of the facial electromyogram signals are extracted from the candidate signal components; The brain electrical response latency and the facial electromyographic activation latency are taken as specific neurophysiological markers.

6. The virtual reality-based anesthetic pain prediction method of claim 5, wherein, Step S4 comprises: The time sequence relationship between the brain electrical response latency and the facial electromyographic activation latency is determined by comparing the time sequence; And based on the continuous changes of the electroencephalogram signals, facial electromyogram signals and galvanic skin response signals during the virtual reality stimulation sequence, signal change trajectories of the time points corresponding to the brain electrical response latency and the facial electromyographic activation latency are extracted; The statistical consistency between the electroencephalogram signal change trajectory, the facial electromyogram signal change trajectory and the galvanic skin response signal change trajectory is calculated to evaluate the dynamic change trajectory consistency.

7. The virtual reality-based anesthetic pain prediction method of claim 6, wherein, Step S5 comprises: When the brain electrical response latency is earlier than the facial electromyographic activation latency, and the statistical consistency between the electroencephalogram signal change trajectory, the facial electromyogram signal change trajectory and the galvanic skin response signal change trajectory exceeds the preset correlation threshold, the corresponding specific neurophysiological marker is determined as a high-specificity pain neuro-marker.

8. The virtual reality-based anesthetic pain prediction method of claim 7, wherein, Step S6 comprises: The time difference between the brain electrical response latency and the facial electromyographic activation latency corresponding to the high-specificity pain neuro-marker, and the statistical consistency between the electroencephalogram signal change trajectory and the galvanic skin response signal change trajectory are taken as input features; The prediction score of the patient's pain during the operation is calculated by using the pre-established pain prediction regression model, and the prediction score is output.

9. The virtual reality-based anesthetic pain prediction method of claim 8, wherein, The pre-established pain prediction regression model is obtained by using the time difference between the brain electrical response latency and the facial electromyographic activation latency corresponding to the high-specificity pain neuro-marker of a plurality of patients in the historical data set, and the statistical consistency between the electroencephalogram signal change trajectory and the galvanic skin response signal change trajectory as input features, and the corresponding true pain score as output label, and is obtained by regression analysis training.

10. A virtual reality based anesthetic pain prediction system for implementing the virtual reality based anesthetic pain prediction method of any one of claims 1-9, characterized in that, Comprise: The application acquisition module is configured to apply a predefined virtual reality stimulation sequence to the patient through the virtual reality device under the condition that the patient receives anesthetic drugs, and synchronously acquire physiological signal data of the patient; The signal screening module is configured to analyze the change direction of a plurality of types of physiological signals in the physiological signal data before and after the virtual reality stimulation sequence, and screen signal components conforming to the predefined cooperative response mode from the physiological signal data as candidate signal components; The marker extraction module is configured to extract neuro-electrophysiological features related to pain perception from the candidate signal components as specific neurophysiological markers; The marker analysis module is configured to analyze the time sequence relationship between the brain electrical response latency and the facial electromyographic activation latency corresponding to the specific neurophysiological markers, and analyze the dynamic change trajectory consistency of the specific neurophysiological markers and a plurality of physiological signals; The condition determination module is configured to determine a high-specificity pain neuro-marker when the brain electrical response latency is earlier than the facial electromyographic activation latency, and the dynamic change trajectory consistency meets the preset condition; The score output module is configured to calculate and output the prediction score of the patient's pain during the operation based on the high-specificity pain neuro-marker.

Images