An esketamine anaesthetic state monitoring method and system

Abstract

The present application provides a ketamine anesthesia state monitoring method and system, comprising: decomposing the collected electroencephalogram signal into multiple frequency bands; extracting the wPLI feature of each frequency band under each lead pair and the PE feature of each channel and each frequency band, respectively; according to the pharmacological mechanism of ketamine, performing specific weighting processing on the wPLI feature and the PE feature of each frequency band to obtain the weighted wPLI feature and the weighted PE feature, which constitute a combined feature vector input into a classification model to output the current anesthesia state label and the confidence. The present application is based on the joint representation of multi-band electroencephalogram functional connectivity features (wPLI) and nonlinear complexity features (PE), and the real-time anesthesia state discrimination of specific weighting processing on the mechanism-related frequency bands such as Theta, which can realize the continuous monitoring of ketamine anesthesia / sedation state with more pertinence, interpretability and stability.

Description

Technical Field

[0001] This invention relates to the field of medical technology, and more specifically, to a method and system for monitoring esketamine anesthesia. Background Technology

[0002] Incomplete postoperative recovery of consciousness and intraoperative awareness are two core risks in the management of general anesthesia. Accurate, continuous, and interpretable monitoring of anesthetic status is crucial for optimizing dosing regimens, ensuring patient safety, and promoting rapid postoperative recovery. Electroencephalography (EEG), as an objective signal that directly reflects the activity of the central nervous system, has been widely used for monitoring the depth of anesthesia. Currently, the most widely used monitoring equipment in clinical practice is the anesthesia depth monitor based on the bispectral index (BIS). However, the BIS algorithm is mainly based on data from traditional anesthetic drugs such as propofol and sevoflurane, which primarily inhibit GABAergic activity. Its feature extraction logic and parameter weights are constructed around typical patterns such as "overall slowing of EEG, enhancement of low frequencies, and decrease in complexity." When the central mechanism of action of the monitored drug differs significantly from that of the aforementioned traditional drugs, the adaptability and accuracy of BIS and similar fixed-model monitoring methods are significantly limited.

[0003] Esketamine, the dextrorotatory isomer of ketamine, is a novel intravenous anesthetic / sedative drug that has attracted considerable attention in recent years. Its core mechanism of action is not simply enhanced GABAergic inhibition, but primarily manifests as antagonism of NMDA receptors, accompanied by remodeling of the thalamic-cortical network, frontal lobe functional connectivity, and specific frequency band neural oscillation patterns. Unlike the purely inhibitory EEG induced by traditional anesthetics, esketamine can induce more complex EEG changes in sedation, anesthesia, and even unresponsive states, such as enhanced activity in specific frequency bands, especially the Theta band, remodeling of phase synchronization relationships between brain regions, and phased changes in EEG complexity. This means that relying solely on traditional empirical power spectrum indicators or fixed frequency band ratios often fails to accurately reflect the evolution of brain functional states under the influence of esketamine. Summary of the Invention

[0004] This invention addresses the technical problems existing in the prior art by providing a method and system for monitoring the anesthetic state of esketamine, thereby solving the problems in the prior art.

[0005] Based on a first aspect of the present invention, a method for monitoring esketamine anesthesia status is provided, comprising: Step S1: Under esketamine anesthesia, continuously collect multi-channel EEG signals from the subject; Step S2: Decompose the EEG signals of each channel into five standard frequency bands: delta, theta, alpha, beta, and gamma, to obtain the brain electron band signals of each channel in each frequency band. Step S3: Based on the sliding time window, calculate the wPLI features of the brain electron band signal in each frequency band of each lead pair for multiple predefined lead pairs, and perform specific weighting processing on the wPLI features of each frequency band according to the pharmacological mechanism of esketamine to obtain the weighted wPLI features of each lead pair. Step S4: Based on the sliding time window, calculate the PE characteristics of the brain electron band signal in each frequency band under each channel, and according to the pharmacological mechanism of esketamine, perform weighted processing on the PE characteristics of each frequency band to obtain the weighted PE characteristics of each channel. Step S5: Concatenate the weighted wPLI features of the multi-lead pairs and the weighted PE features of the multi-channel pairs to construct a combined feature vector; Step S6: Input the combined feature vector into the classification model and output the current anesthesia status label and confidence level.

[0006] Based on a second aspect of the present invention, an esketamine anesthesia status monitoring system is provided, comprising: The EEG signal acquisition module is used to continuously acquire multi-channel EEG signals from subjects under esketamine anesthesia. The data processing and control unit is used to decompose the EEG signals of each channel into five standard frequency bands: delta, theta, alpha, beta, and gamma, to obtain the brain electron band signals of each channel in each frequency band. Based on a sliding time window, for multiple predefined lead pairs, the wPLI features of the brain electron band signals in each frequency band of each lead pair are calculated, and the wPLI features of each frequency band are specifically weighted according to the pharmacological mechanism of esketamine to obtain the weighted wPLI features of each lead pair. It is also used to calculate the PE features of the brain electron band signals in each frequency band of each channel based on the sliding time window, and to weight the PE features of each frequency band according to the pharmacological mechanism of esketamine to obtain the weighted PE features of each channel. The weighted wPLI features of multiple lead pairs and the weighted PE features of multiple channels are concatenated to construct a combined feature vector. Finally, the combined feature vector is input into a classification model to output the current anesthesia state label and confidence level.

[0007] This invention provides a method and system for monitoring esketamine anesthesia status. Based on the pharmacological characteristics of esketamine, it uses multi-band electroencephalogram functional connectivity features (wPLI) and nonlinear complexity features (PE) for joint characterization, and performs specific weighting processing on mechanism-related frequency bands such as Theta for real-time anesthesia status discrimination. This enables more targeted, interpretable, and stable continuous monitoring of esketamine anesthesia / sedation status, providing more effective technical support for the clinical application and research evaluation of this type of drug. Attached Figure Description

[0008] Figure 1 A flowchart of a method for monitoring esketamine anesthesia status according to an embodiment of the present invention; Figure 2 This is an overall flowchart of a method for monitoring the anesthesia state of esketamine provided in one embodiment of the present invention; Figure 3 This is a sub-flowchart of closed-loop adaptive parameter tuning in an esketamine anesthesia monitoring method according to an embodiment of the present invention; Figure 4 This is a structural block diagram of an esketamine anesthesia monitoring system provided in one embodiment of the present invention. Detailed Implementation

[0009] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. In addition, the technical features of the various embodiments or individual embodiments provided by the present invention can be arbitrarily combined with each other to form feasible technical solutions. Such combinations are not constrained by the order of steps and / or structural composition patterns, but must be based on the ability of those skilled in the art to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be considered that such a combination of technical solutions does not exist and is not within the scope of protection claimed by the present invention.

[0010] Based on the above situation, this invention does not follow the fixed feature-fixed weight discrimination approach for general anesthetic drugs. Instead, it focuses on the pharmacological characteristics of esketamine and constructs a real-time anesthesia state discrimination method based on the joint representation of multi-band EEG functional connectivity features (wPLI) and nonlinear complexity features (PE), and specifically weights the frequency bands related to mechanisms such as Theta. This method further combines a pharmacologically adaptive weighted Bayesian model, consistency verification, and a closed-loop adaptive parameter tuning mechanism. Compared with ordinary Bayesian models, the improvements to the model in this invention include at least the following: the input uses a combination of wPLI and PE features weighted by pharmacological mechanisms instead of the general original features; the training uses an esketamine-specific dataset labeled with MOAA / S scores to estimate the class prior and conditional probability distribution; the probability density estimation preferably uses kernel density estimation to improve the adaptability to non-Gaussian feature distributions; the output simultaneously provides the state label, posterior probability, and classification confidence; and the execution is linked with the closed-loop adaptive parameter tuning mechanism to perform constrained updates to the weights of key frequency bands when the confidence is low or the state is inconsistent. This enables more targeted, interpretable, and stable continuous monitoring of esketamine anesthesia / sedation, providing more effective technical support for the clinical application and research evaluation of this type of drug.

[0011] Figure 1 A flowchart of a method for monitoring the anesthesia status of esketamine according to an embodiment of the present invention is shown, as follows: Figure 1 and Figure 2 As shown, the method includes the following steps: Step S1: Under esketamine anesthesia, continuously collect multi-channel EEG signals from the subject.

[0012] Understandably, the raw EEG signals of the subject under esketamine anesthesia are first collected. In this embodiment of the invention, four-channel surface electrodes are used, arranged according to the international 10-20 system at positions Fp1, Fp2, F7, and F8. Raw EEG signals from the frontal / frontotemporal region of the subject are continuously collected at a sampling rate of 500 Hz across the four channels (Fp1, Fp2, F7, and F8). The collected raw EEG signals are then written into a buffer in chronological order. Specifically, the raw EEG signals are written into a circular buffer in 24-bit signed integer format. Let the coverage duration of the circular buffer be T seconds, the number of channels be C, and the sampling rate be Fs, then the total number of data points N in a single circular buffer is:

[0013] In this embodiment, C=4, Fs=500 Hz, T=10 s, therefore N=20000 data points. This circular buffer provides continuous data input for the subsequent 10-second feature window calculation.

[0014] After acquiring the raw EEG signals, the acquired raw EEG signals are preprocessed, and automatic artifact detection is performed by combining amplitude threshold, gradient threshold and continuity criteria to mark or remove abnormal signal segments.

[0015] The preprocessing mainly includes digital bandpass filtering of the raw EEG signal at 0.5–45 Hz and notch filtering at 50 Hz to suppress low-frequency drift, high-frequency noise, and power frequency interference. A fourth-order IIR digital filter can be used for the bandpass filter, and its discrete-domain transfer function is written as:

[0016] in, For unit delay operators, - These are the filter numerator coefficients. - These are the denominator coefficients of the filter.

[0017] Simultaneously, automatic artifact detection is performed on the EEG signals. Let... This represents the amplitude at the nth sampling point. =1 / Fs represents the sampling interval. Therefore, if any of the following conditions are met within any 250ms detection window, the EEG signal within that window is determined to be an artifact segment. The conditions are: or

[0018] in, The amplitude threshold is preferably 100 µV; The gradient threshold is preferably 50 µV / ms. EEG signal segments marked as artifacts are not included in the effective feature calculation of the current 10-second feature window.

[0019] Step S2: Decompose the EEG signals of each channel into five standard frequency bands: delta, theta, alpha, beta, and gamma, to obtain the brain electron band signals of each channel in each frequency band.

[0020] Understandably, this step decomposes the preprocessed EEG signal into five standard frequency bands: delta (0.5-4 Hz), theta (4-8 Hz), alpha (8-13 Hz), beta (13-30 Hz), and gamma (30-45 Hz), obtaining the brain electron band signal for each channel in each frequency band. Considering the strong correlation between the pharmacological effects of esketamine and the activity in the Theta band, a higher weight is applied to the Theta band-related features in the subsequent steps S3 (calculation of EEG functional connectivity wPLI features) and S4 (calculation of nonlinear complexity PE features).

[0021] Step S3: Based on the sliding time window, calculate the wPLI features of the brain electron band signal in each frequency band for each of the multiple predefined lead pairs. Based on the pharmacological mechanism of esketamine, perform specific weighting on the wPLI features of each frequency band to obtain the weighted wPLI features of each lead pair.

[0022] Understandably, for the decomposed brain electron band signals of each frequency band, a sliding time window and sliding step size are set. In this embodiment of the invention, within a 10-second sliding time window, with a step size of 2 seconds, the wPLI features of each frequency band signal are calculated for multiple predefined lead pairs. The multiple predefined lead pairs can be at least a portion of Fp1-F7, Fp1-F8, Fp2-F7, Fp2-F8, Fp1-Fp2, and F7-F8. Preferably, the instantaneous phase of the brain electron band signals of each frequency band is first obtained based on the Hilbert transform, and then the corresponding wPLI feature values ​​are calculated based on the phase difference sequence. Subsequently, based on the pharmacological mechanism of esketamine, the wPLI features of different frequency bands are weighted, preferably assigning a higher weight to the wPLI features corresponding to the Theta band, to obtain the weighted wPLI features of each lead pair, thus forming a weighted functional connectivity feature set across leads and frequency bands.

[0023] Specifically, within the current sliding time window, for any lead pair and any frequency band The brain electron band signal is first analyzed by Hilbert transform to obtain the analytic signal and extract the instantaneous phase, and then recorded as leads. and leads In frequency band Next, time The instantaneous phases are respectively and Then the phase difference sequence Represented as:

[0024] Further based on the phase difference sequence Calculate the corresponding cross-spectral imaginary part sequence ,in, Indicates lead With leads In frequency band ,time The mutual spectrum.

[0025] Among them, mutual spectrum The imaginary part can be directly derived from the phase difference, cross spectrum The physical meaning is: Its imaginary part is: ,in, The cross-spectral amplitude represents the coherence strength of the two signals at that frequency point. It is directly determined by the phase difference and reflects the direction of phase lag.

[0026] The original weighted phase lag index wPLI characteristic can be expressed as:

[0027] in, Indicates lead pair and frequency band wPLI features This indicates averaging the samples within the current 10-second sliding time window, with Im{·} representing the imaginary part. Subsequently, based on the pharmacological mechanism of esketamine, corresponding weight vectors are set for the wPLI features of each frequency band, and specific weighting processing is performed on the wPLI features of each frequency band to obtain the pair for each lead. Weighted wPLI features:

[0028] in, Indicates the frequency band corresponding to the wPLI feature. The pharmacological weighting coefficient, Indicates lead pair The weighted wPLI features; preferably, the theta band corresponding to Take 2.0, the corresponding frequency band of delta Take 1.5, the corresponding frequencies for alpha, beta, and gamma bands Set the value to 1.0. Calculate the weighted wPLI feature for each lead pair, thereby obtaining the set of weighted functional connectivity features across lead pairs and frequency bands.

[0029] Step S4: Based on the sliding time window, calculate the PE characteristics of the brain electron band signal in each frequency band under each channel, and according to the pharmacological mechanism of esketamine, perform weighted processing on the PE characteristics of each frequency band to obtain the weighted PE characteristics of each channel.

[0030] Understandably, within the same 10-second sliding time window as in step S3, the permutation entropy of the brain electron band signals for each channel and frequency band is calculated separately. By reconstructing the phase space of the time series and statistically analyzing the probability of different permutation patterns, PE feature values ​​characterizing the signal complexity of each channel and frequency band are obtained. Subsequently, based on the pharmacological mechanism of esketamine, the PE features of different frequency bands are weighted, preferably assigning higher weights to the PE features corresponding to the Theta band, to obtain the weighted PE features of each channel, forming a weighted complexity feature set across channels and frequency bands.

[0031] Specifically, regarding channels frequency band EEG time series By embedding dimension and time delay Construct phase space vectors:

[0032] right The elements are mapped to a permutation pattern according to their size. Statistics of all Permutation patterns, permutation patterns The relative frequency of occurrence is Then the channel frequency band The permutation entropy is defined as:

[0033] In this embodiment, = 3, = 1, = 6. (The rest of the text appears to be a list of numbers and symbols, possibly related to a document or instruction.) As a channel frequency band PE characteristics.

[0034] Subsequently, based on the pharmacological mechanism of action of esketamine, weighted vectors were set for the PE features of each frequency band. Based on the weighted vectors, specific weighting processing was performed on the PE features of each frequency band to obtain the weighted PE features:

[0035] in, Indicates the frequency band corresponding to PE characteristics The pharmacological weighting coefficient, This represents the weighted PE characteristics of channel C; preferably, the theta band corresponds to... Take 2.0, the corresponding frequency band of delta Take 1.5, the corresponding frequencies for alpha, beta, and gamma bands We set it to 1.0. This yields a weighted complexity feature set across channels and frequency bands, which includes the weighted PE features of each channel.

[0036] Step S5: Concatenate the weighted wPLI features of the multi-lead pair and the weighted PE features of the multi-channel pair to construct a combined feature vector.

[0037] Understandably, the multi-lead pairs and multi-band weighted wPLI features obtained in step S3 within the current sliding time window are concatenated with the channel-specific and frequency-band weighted PE features obtained in step S4 to construct the combined feature vector for the input classification model. ,Right now:

[0038] in, This represents the combined feature vector input to the classification model. This represents the weighted wPLI feature set output in step S3. This represents the weighted PE feature set output in step S4. Step S5 no longer performs weight assignment separately, but instead performs unified concatenation and organization of the features that have already undergone weighting.

[0039] Step S6: Input the combined feature vector into the classification model and output the current anesthesia status label and confidence level.

[0040] Understandably, the combined feature vector formed by the specific weighting process completed in steps S3 and S4 is input into the pre-trained pharmacological adaptive weighted Bayesian model. Compared to ordinary Bayesian classification models, the pharmacologically adaptive weighted Bayesian model has at least the following improvements: First, the input features no longer use general raw EEG statistics, but instead use a combined feature vector composed of multi-lead pairs, multi-band wPLI features and multi-channel, multi-band PE features weighted by the pharmacological mechanism of esketamine; Second, the pharmacologically adaptive weighted Bayesian model is trained based on the esketamine anesthesia dataset labeled with MOAA / S scores, and establishes class prior probabilities and conditional probability distributions for the three types of anesthesia states; Third, the probability density of each feature under class conditions is preferably fitted using kernel density estimation to improve the adaptability to complex distribution features; Fourth, the output of the pharmacologically adaptive weighted Bayesian model includes not only the anesthesia state category of the current sliding time window, but also the corresponding posterior probability and the instantaneous confidence level represented by the maximum posterior probability; Fifth, the pharmacologically adaptive weighted Bayesian model can be linked with the closed-loop adaptive parameter tuning mechanism in subsequent steps, and when the maximum posterior probability (confidence level) is continuously lower than the threshold or the results of consecutive windows are inconsistent, the discrimination is re-executed based on the updated key frequency band weights. The current anesthesia status is labeled according to the category corresponding to the maximum posterior probability, where 0, 1, and 2 correspond to conscious / mild sedation, moderate sedation, and deep sedation / unresponsiveness, respectively.

[0041] Specifically, the feature vector V obtained in step S5 is input into the pre-trained pharmacological adaptive weighted Bayesian model. Let the combined feature vector of the samples to be classified be... D is the dimension of the combined feature vector, and the category to which the sample to be classified belongs. The posterior probability can be expressed as:

[0042] in, The combined feature vector V belongs to the category The posterior probability, For the i-th feature in category The conditional probability density is obtained under the given conditions, and kernel density estimation is preferably used for fitting the conditional probability density. The kernel density estimation uses a Gaussian kernel function:

[0043] ≈

[0044] in, For the first Number of training samples per class For the first The bandwidth parameter of each feature For the first Class 1 The sample at the th The values ​​that can be taken on each feature The preferred kernel function is a Gaussian kernel.

[0045] Kernel density estimation is preferred for estimating the conditional probability density of each feature under each class condition. A fit is performed, and the fitting result is then substituted into the Bayesian posterior probability formula to calculate the posterior probability that the sample to be classified belongs to one of the three anesthesia categories. The category corresponding to the maximum posterior probability is then used as the current anesthesia output state. Here, category 0 corresponds to awake / mild sedation, category 1 corresponds to moderate sedation, and category 2 corresponds to deep sedation / unresponsiveness. The maximum posterior probability value is also used as the instantaneous confidence level at the current moment.

[0046] After classifying the anesthesia status in the current sliding time window in step S6, step S7 is also included: monitoring the consistency of the anesthesia status classification results within consecutive sliding time windows.

[0047] Preferably, N=10 consecutive sliding time windows are set as consistency judgment intervals; when the anesthesia state labels of consecutive sliding time windows are consistent, the current anesthesia state is confirmed to be valid; when multiple consecutive sliding time windows show inconsistent anesthesia states, or when the maximum posterior probability of multiple sliding time windows is consistently lower than the preset reliability threshold Tc, a closed-loop adaptive parameter tuning mechanism is triggered to adjust the weight vectors of wPLI features and PE features in each frequency band. The preferred value is 0.80.

[0048] See Figure 3 During parameter tuning, the feature weight vector is updated under constraints according to the preset pharmacological rule library and error signal type. During the update process, the weights related to the Theta band are corrected first, and the updated weight vector is normalized before entering the next round of "monitoring-evaluation-adjustment-verification" cycle to improve the adaptability under individual differences and interference conditions.

[0049] Specifically, during each parameter tuning, only the weight vector of the Theta band is adjusted, while the weight vectors of other bands remain unchanged. In this embodiment of the invention, the weight vectors of the wPLI feature and PE feature are the same for each band.

[0050] During parameter tuning, the weight vector currently used in steps S3 and S4 is taken as the center. In this embodiment, except for the weights related to the theta band, the weights corresponding to other frequency bands remain unchanged. In other embodiments, the weights of the delta band or multiple key frequency bands can also be jointly adjusted according to the error signal type. Taking the weights related to the theta band as an example, two candidate weight vectors are constructed:

[0051]

[0052] in, The perturbation step size is preferably 0.05; This represents a unit direction vector that takes a value of 1 only at the relevant feature positions in the theta band and a value of 0 at all other positions. , and The corresponding three sets of weight vectors for each frequency band.

[0053] Three sets of weight vectors are used in the subsequent M sliding time windows. , and Perform anesthesia state classification and calculate the average confidence score corresponding to the classification of each group's weight vector:

[0054] in, Represents the weight vector Average classification confidence over M consecutive sliding time windows; Indicates the first A sliding time window in the weight vector The maximum posterior probability obtained is M, which is preferably 5. The choice is to make... The largest weight vector is used as the new current weight vector, and it is written back to steps S3 and S4 to continue execution. To ensure stability, a non-negativity constraint is applied to the updated weight vector. , Let p be the weight vector of the p-th group (p=1,2,3), which is the updated current weight vector. Through the above closed-loop process of "monitoring-evaluation-adjustment-verification", the system's adaptability under individual differences, artifact interference, and state boundary intervals is improved.

[0055] Throughout the anesthesia monitoring process, the original EEG waveform, trend curves of wPLI and PE features in each frequency band, current anesthesia status label and its confidence information are refreshed every 1 second. The original EEG data, wPLI features, PE features, anesthesia status classification results and timestamps are stored simultaneously for intraoperative monitoring, postoperative traceability and model optimization.

[0056] It should be noted that throughout the entire process of determining the anesthesia state, the classification results of the anesthesia state are continuously visualized and recorded. Preferably, the original EEG waveform, the trend curves of wPLI features and PE features of each frequency band, the current anesthesia state label and its confidence information are refreshed at a 1-second cycle, and the original EEG data, feature values, classification results and timestamps are stored simultaneously for intraoperative monitoring, postoperative traceability and model optimization.

[0057] See Figure 4 This invention provides an esketamine anesthesia monitoring system according to one embodiment of the present invention. The monitoring system comprises the following six core modules, mainly including: The EEG signal acquisition module 401 is used to continuously acquire multi-channel EEG signals from the subject under esketamine anesthesia.

[0058] The EEG signal acquisition module 401 uses medical-grade Ag / AgCl disposable electrodes that comply with the IEC 60601-1 safety standard for medical electrical equipment. The electrode layout employs a simplified frontal lobe lead configuration, specifically: Channel 1 (CH1): Fp1, Channel 2 (CH2): Fp2, Channel 3 (CH3): F7, Channel 4 (CH4): F8, with the reference electrode placed on the right earlobe (A2). Skin preparation is performed before electrode placement to ensure that the impedance of a single electrode is below 5 kΩ. The connecting cable is a 4-core shielded cable with an effective length of 1.5 meters and a 4-pin LEMO connector with locking and anti-misinsertion functions.

[0059] The analog front-end signal conditioning module 402 is used to preprocess the raw EEG signal and detect artifacts.

[0060] The analog front-end signal conditioning module 402 integrates amplification, filtering, and protection circuits. The preamplifier uses a Texas Instruments INA333 instrumentation amplifier with a gain of 1000. The circuit employs right-leg drive (RLD) technology to improve the common-mode rejection ratio to over 120 dB, effectively suppressing 50 Hz power frequency interference. The analog filter uses a two-stage active filter: the first stage is a high-pass filter with a cutoff frequency of 0.5 Hz, used to eliminate DC offset and slow baseline drift; the second stage is a low-pass anti-aliasing filter with a cutoff frequency of 45 Hz; the total passband of the filter is 0.5–45 Hz.

[0061] Analog-to-digital converter module 403 is used to convert analog EEG signals into digital EEG signals.

[0062] The analog-to-digital conversion module 403 uses Analog Devices' ADAS1000-4 low-power 4-channel analog front-end chip. This system configures it for EEG acquisition mode, with a sampling rate of 500 Hz and an input reference noise of 1.0 µVpp. Each channel has a built-in 24-bit Sigma-Delta ADC. The chip communicates with the control unit via an SPI interface, continuously outputting 4 channels of raw EEG data.

[0063] The data processing control unit 404 is used to process EEG signals, including extracting wPLI features of different lead pairs and different frequency bands, PE features of different channels and different frequency bands, calculating the weighted wPLI features of each lead pair and the weighted PE features of each channel, and identifying the subject's current anesthesia state based on the weighted wPLI features of each lead pair and the weighted PE features of each channel.

[0064] The core processor of the data processing and control unit 404 adopts an ARM Cortex-A53 architecture processor. This unit is responsible for tasks such as data caching, digital filtering, frequency band decomposition, wPLI and PE calculation, feature vector construction, pharmacological adaptive weighted Bayesian model inference, consistency verification, and closed-loop parameter tuning. The system runs a real-time operating system. High-priority tasks are responsible for sampling and caching, medium-priority tasks are responsible for digital preprocessing, and algorithm tasks are scheduled to refresh and update their status every 1 second.

[0065] The data transmission module 405 is used for data transmission.

[0066] The data transmission module 405 provides both wired and wireless data transmission methods. The wired interface uses a standard USB 3.0 for connecting to a host computer for data recording and in-depth analysis. Wireless transmission supports Wi-Fi, and the wireless link uses AES-128 encryption, enabling the push of real-time anesthesia status, characteristic curves, and alarm information to the central monitoring system or mobile terminal.

[0067] The display and interaction module 406 is used to display the original EEG signal, the trend curves of wPLI and PE features of each frequency band, the current anesthesia status label and the corresponding confidence level according to a preset refresh cycle.

[0068] The display and interaction module 406 uses a 10.1-inch medical display screen. The user interface is divided into four main areas: a waveform display area for real-time scrolling display of raw EEG signals from four channels; a trend graph area for displaying the change curves of wPLI and PE over the past 5 minutes; a status indicator area for displaying the current anesthesia status and confidence level; and a system status area for displaying electrode impedance, signal quality, and network connection status.

[0069] This invention provides a method and system for monitoring esketamine anesthesia status. Based on the esketamine anesthesia monitoring scenario, it constructs a complete technical solution encompassing "multi-band EEG feature extraction—pharmacological mechanism-specific weighted processing—pharmacological adaptive weighted Bayesian model discrimination—consistency confirmation and closed-loop parameter tuning—result display and data recording." This solution is not a simple transplantation of general anesthesia depth monitoring methods, but is specifically designed for the characteristics of esketamine's effects, primarily through NMDA receptor antagonism, accompanied by enhanced Theta band activity, reconstructed brain region connectivity, and dynamic changes in EEG complexity. Compared to existing monitoring methods based on fixed frequency band power ratios or general empirical indices, this invention has at least the following advantages: 1. A characteristic system was constructed based on the pharmacological mechanism of esketamine, which has stronger specificity.

[0070] Based on full-band decomposition, this invention simultaneously extracts wPLI features from multiple lead pairs and multiple frequency bands, as well as PE features from each channel and each frequency band. It jointly characterizes the esketamine anesthesia state from two dimensions: "functional connectivity" and "nonlinear complexity," making the monitoring indicators more closely match the core neurophysiological effects of the drug.

[0071] 2. Applying specific weighting to the relevant frequency bands can highlight key discrimination information.

[0072] To address the discriminative value of the Theta band activity under esketamine influence, this invention employs pharmacological mechanism-guided weighting of relevant frequency band features during wPLI and PE calculations, rather than a uniform and coarse approach in the general classification stage. This design enhances the contribution of key frequency band features to overall discrimination, improving the ability to differentiate between different states such as conscious / mildly sedated, moderately sedated, and deeply sedated / unresponsive.

[0073] 3. By combining connectivity and complexity features, the completeness of state representation can be improved.

[0074] Traditional methods often rely on a single power spectrum index or a single comprehensive index, which makes it difficult to simultaneously reflect both "whether there are synchronous changes between brain regions" and "whether the complex structure is still maintained within a single channel signal." This invention incorporates wPLI and PE into the same discriminant framework, which can simultaneously reflect changes in brain network organization and the complexity of neural activity during esketamine administration, thereby improving the comprehensiveness and robustness of state assessment.

[0075] 4. Achieving differentiated state discrimination for esketamine through a pharmacologically adaptive weighted Bayesian model.

[0076] This invention employs a pharmacologically adaptive weighted Bayesian model to output the posterior probability of each anesthetic state, rather than directly applying a conventional Bayesian classification model. Compared to traditional Bayesian models, this invention makes at least the following improvements: First, the model input consists of a combination of wPLI and PE features weighted specifically by pharmacological mechanisms in steps S3 and S4; second, the model training is geared towards the esketamine scenario, estimating the class prior and conditional probability density based on a dedicated dataset labeled with MOAA / S scores; third, the conditional probability density preferably uses kernel density estimation for fitting, enhancing its adaptability to nonlinear and non-Gaussian distribution features; fourth, the model output not only provides the state class but also simultaneously provides the posterior probability and instantaneous confidence level; fifth, the model is linked to the closed-loop parameter tuning mechanism in step S7, allowing for re-discrimination after updating the weights of key frequency bands under low confidence conditions. This design facilitates physicians in identifying boundary states, low-confidence states, and state transition processes, and also benefits researchers in conducting pharmacodynamic analysis and model performance evaluation.

[0077] 5. It has a consistency confirmation and closed-loop adaptive parameter tuning mechanism, which improves the stability of continuous monitoring.

[0078] When inconsistencies occur in the states of consecutive windows or the classification confidence remains consistently low, the system can automatically trigger closed-loop parameter tuning, performing constrained updates and re-verification of the feature weights in steps S3 and S4. This mechanism can dynamically correct for individual differences, artifact interference, and state transition periods, reducing the impact of occasional fluctuations on the final output and improving the monitoring stability and adaptability in complex scenarios.

[0079] 6. The output information is rich in dimensions, suitable for both clinical monitoring and new drug research.

[0080] This invention not only outputs the anesthesia status category, but also simultaneously displays the original EEG waveform, the trends of wPLI and PE changes in each frequency band, and the classification confidence level, while recording timestamps and feature data. This design facilitates clinicians' understanding of the criteria for status determination and also facilitates research on the mechanism of action of esketamine, postoperative retrospective analysis, and model iterative optimization.

[0081] 7. It can be implemented using conventional EEG acquisition hardware, and has good engineering feasibility and promotional value.

[0082] This invention can be implemented based on 4-channel frontal / frontotemporal EEG acquisition hardware, standard analog-to-digital conversion unit and ARM Cortex-A53 processing platform. It does not rely on specially customized sensors or highly complex experimental platforms, making it easy to deploy as a standalone device or integrate as an algorithm module into existing anesthesia monitoring systems. It has clear clinical translation prospects.

[0083] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0084] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0085] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0086] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0087] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0088] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0089] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for monitoring the state of anesthesia with esketamine, characterized in that, include: Step S1: Under esketamine anesthesia, continuously collect multi-channel EEG signals from the subject; Step S2: Decompose the EEG signals of each channel into five standard frequency bands: delta, theta, alpha, beta, and gamma, to obtain the brain electron band signals of each channel in each frequency band. Step S3: Based on the sliding time window, calculate the wPLI features of the brain electron band signal in each frequency band of each lead pair for multiple predefined lead pairs, and perform specific weighting processing on the wPLI features of each frequency band according to the pharmacological mechanism of esketamine to obtain the weighted wPLI features of each lead pair. Step S4: Based on the sliding time window, calculate the PE characteristics of the brain electron band signal in each frequency band under each channel, and according to the pharmacological mechanism of esketamine, perform weighted processing on the PE characteristics of each frequency band to obtain the weighted PE characteristics of each channel. Step S5: Concatenate the weighted wPLI features of the multi-lead pairs and the weighted PE features of the multi-channel pairs to construct a combined feature vector; Step S6: Input the combined feature vector into the classification model and output the current anesthesia status label and confidence level.

2. The method for monitoring esketamine anesthesia status according to claim 1, characterized in that, In step S1, under esketamine anesthesia, multi-channel electroencephalogram (EEG) signals are continuously collected from the subject, including: Raw EEG signals from four channels (Fp1, Fp2, F7, and F8) of the subject were continuously acquired at a fixed sampling frequency. These raw EEG signals were then written into a circular buffer. Let the coverage duration of the circular buffer be T seconds, the number of channels be C, and the sampling rate be Fs. Then, the total number of data points N in a single circular buffer is: 。 3. The method for monitoring esketamine anesthesia status according to claim 1, characterized in that, Step S1, under esketamine anesthesia, involves continuously acquiring multi-channel EEG signals from the subject, followed by: The four-channel raw EEG signals are preprocessed, including digital bandpass filtering and notch filtering. Specifically, a fourth-order IIR digital filter is used to perform digital bandpass filtering on the four-channel raw EEG signals. The discrete-domain transfer function expression of the fourth-order IIR digital filter is as follows: in, For unit delay operators, - These are the filter numerator coefficients. - These are the denominator coefficients of the filter; set up This represents the amplitude of the EEG signal at the nth sampling point. =1 / Fs represents the sampling interval, and Fs represents the sampling frequency. Therefore, when the judgment condition is met within any detection window h, the EEG signal within the detection window h is determined to be a artifact segment. The judgment condition is: or in, For a detection window, For amplitude threshold, Gradient threshold; EEG signals identified as artifacts were removed.

4. The method for monitoring esketamine anesthesia status according to claim 1, characterized in that, Step S3, based on a sliding time window, calculates the wPLI features of the brain electron band signal in each frequency band for multiple predefined lead pairs, and performs specific weighting processing on the wPLI features of each frequency band according to the pharmacological mechanism of esketamine, to obtain the weighted wPLI features of each lead pair, including: Set a sliding time window and predefine multiple lead pairs, wherein the multiple lead pairs include at least a portion of Fp1–F7, Fp1–F8, Fp2–F7, Fp2–F8, Fp1–Fp2, and F7–F8; For any lead pair and any frequency band The brain electron band signal was analyzed using Hilbert transform, and the instantaneous phase was extracted from the analyzed signal and recorded as leads. and leads In frequency band Next, time The instantaneous phases are respectively and Then the phase difference sequence Represented as: Based on the phase difference sequence Calculate the corresponding cross-spectral imaginary part sequence ,in, Indicates lead With leads In frequency band ,time The cross spectrum of wPLI is then represented as: ； in, This indicates averaging the samples within the current sliding time window, and Im{·} denotes taking the imaginary part. Indicates lead and leads In frequency band wPLI features below; Based on the pharmacological mechanism of action of esketamine, the wPLI features of each frequency band were specifically weighted to obtain the weighted wPLI features: in, This represents the pharmacological weighting coefficient of the frequency band f corresponding to the wPLI feature. The pharmacological weighting coefficient of the theta frequency band corresponding to the wPLI feature is the largest. Indicates lead pair and frequency band wPLI features Indicates lead pair The weighted wPLI features.

5. The method for monitoring esketamine anesthesia status according to claim 1, characterized in that, Step S4 involves calculating the PE characteristics of the brain electron band signal in each frequency band under each channel based on the sliding time window, and weighting the PE characteristics of each frequency band according to the pharmacological mechanism of esketamine to obtain the weighted PE characteristics of each channel, including: Within the same sliding time window as in step S3, the permutation entropy of the brain electron band signals for each channel and frequency band is calculated. Specifically, for each channel... frequency band The following time series By embedding dimension and time delay Construct a phase space vector, which is represented as: right Mapping the elements according to their size into various permutation patterns Statistics of all There are several permutation patterns, each of which... The relative frequency of occurrence is Then the channel frequency band permutation entropy Defined as: permutation entropy As a channel frequency band PE characteristics; Based on the pharmacological mechanism of action of esketamine, the PE characteristics of each frequency band in each channel were specifically weighted to obtain the weighted PE characteristics of each channel: in, Indicates the frequency band corresponding to PE characteristics The pharmacological weighting coefficient, For channel The weighted PE characteristics.

6. The method for monitoring esketamine anesthesia status according to claim 1, characterized in that, Step S6 involves inputting the combined feature vector into a classification model and outputting the current anesthesia state label and confidence level, including: The combined feature vector is input into a pre-trained pharmacological adaptive weighted Bayesian model, which outputs the current anesthesia state label and confidence level. Wherein, the combined feature vector of the sample to be classified represents V D is the dimension of the combined feature vector, which belongs to the category. The posterior probability is expressed as: in, The sample to be classified is a category The posterior probability, k=0,1,2, For the i-th feature vector in category The conditional probability density under the given conditions is fitted using kernel density estimation. The posterior probabilities of the three categories are calculated respectively, and the category corresponding to the maximum posterior probability is used as the label of the current anesthesia state. The maximum posterior probability value is used as the instantaneous confidence at the current moment. Category 0 corresponds to awake / mild sedation, category 1 corresponds to moderate sedation, and category 2 corresponds to deep sedation / unresponsiveness.

7. The method for monitoring esketamine anesthesia status according to claim 1, characterized in that, Step S6 involves inputting the combined feature vector into the classification model, outputting the current anesthesia state label and confidence level, and then further includes: Obtain the classification results of anesthesia status through continuous sliding time windows, and set N consecutive sliding time windows as the consistency judgment interval; When the anesthesia status classification results are consistent across N consecutive sliding time windows, the current anesthesia status label is determined to be valid. When the classification results of anesthesia status are inconsistent in consecutive sliding time windows, or when the confidence of multiple sliding time windows is consistently lower than the preset threshold Tc, closed-loop adaptive parameter tuning is triggered to adjust the weight vectors of wPLI features and PE features in each frequency band. The weight vectors of wPLI features and PE features in each frequency band are the same.

8. The method for monitoring esketamine anesthesia status according to claim 7, characterized in that, The adjustment of the weight vectors for the wPLI features and the weight vectors for the PE features of each frequency band includes: Using the current weight vector as the center, construct two candidate weight vectors: in, The perturbation step size; This represents a unit direction vector that takes a value of 1 only at the feature positions in the theta band and a value of 0 at the feature positions in the other frequency bands. This is the current weight vector. and There are two candidate weight vectors; Weight vector based on theta band , and Obtain the three sets of weight vectors corresponding to each frequency band; Anesthesia status is classified based on each set of weight vectors within M consecutive sliding time windows, and the average confidence level corresponding to each set of weight vectors is calculated. The average confidence level is expressed as: in, Represents the weight vector Average classification confidence over M consecutive sliding time windows; Indicates the first A window in the weight vector The maximum posterior probability obtained; Choose to The largest weight vector updates the current weight vector, and a nonnegativity constraint is applied to the updated current weight vector. , Let p represent the weight vector of the p-th group, which is the updated current weight vector. Write the updated current weight vector back to steps S3 and S4 and continue execution.

9. The method for monitoring esketamine anesthesia status according to claim 1, characterized in that, Step S6 involves inputting the combined feature vector into the classification model, outputting the current anesthesia state label and confidence level, and then further includes: The system displays the original EEG signal, trend curves of wPLI and PE features in each frequency band, current anesthesia status label and corresponding confidence level according to the preset refresh cycle.

10. A system for monitoring the state of anesthesia with esketamine, characterized in that, include: The EEG signal acquisition module is used to continuously acquire multi-channel EEG signals from subjects under esketamine anesthesia. The data processing and control unit is used to decompose the EEG signals of each channel into five standard frequency bands: delta, theta, alpha, beta, and gamma, to obtain the brain electron band signals of each channel in each frequency band. Based on a sliding time window, for multiple predefined lead pairs, the wPLI features of the brain electron band signals in each frequency band of each lead pair are calculated, and the wPLI features of each frequency band are specifically weighted according to the pharmacological mechanism of esketamine to obtain the weighted wPLI features of each lead pair. It is also used to calculate the PE features of the brain electron band signals in each frequency band of each channel based on the sliding time window, and to weight the PE features of each frequency band according to the pharmacological mechanism of esketamine to obtain the weighted PE features of each channel. The weighted wPLI features of multiple lead pairs and the weighted PE features of multiple channels are concatenated to construct a combined feature vector. Finally, the combined feature vector is input into a classification model to output the current anesthesia state label and confidence level.

Images