Anesthesia depth monitoring method, system and equipment based on auditory response and medium

Abstract

The invention provides an anesthesia depth monitoring method, system and device based on auditory response and a medium, and the method comprises the following steps: constructing and outputting a composite auditory stimulation sequence formed by alternate circulation of a steady state evoked segment and a cognitive probing segment, the cognitive probing segment being a rapid presentation with a stimulation interval smaller than evoked potential duration; the continuous scalp electroencephalogram of the subject is acquired. Intercepting steady-state evoked section data, extracting steady-state auditory evoked response features and calculating a first anesthesia depth index; intercepting cognitive exploration section data, extracting event-related potential by adopting an overlapped signal separation algorithm, and calculating a second anesthesia depth index. And determining two confidence coefficients weights based on the signal quality of the steady-state section and the residual noise of the cognitive section, carrying out weighted fusion on the two indexes according to the weights, and generating and outputting a comprehensive anesthesia depth index for real-time and robust anesthesia depth evaluation and prompt. By implementing the technical scheme provided by the invention, the problem of low updating rate of traditional cognitive potential monitoring data can be solved.

Description

Technical Field

[0001] This application relates to the field of clinical anesthesia and neurophysiological monitoring technology, and in particular to a method, system, device and medium for monitoring the depth of anesthesia based on auditory response. Background Technology

[0002] With the rapid development of precision medicine and perioperative medicine, the number of general anesthesia surgeries continues to rise, and the clinical requirements for anesthesia quality management are constantly increasing. How to accurately, in real time and with interference resistance monitor the patient's level of consciousness, sedation and cognitive function during surgery has become a core requirement for ensuring patient safety and preventing intraoperative awareness or excessive anesthesia.

[0003] In existing technologies, anesthesia depth monitoring is typically achieved through exponential analysis based on spontaneous EEG or single-mode auditory evoked potential (AEP) techniques. For example, the amplitude or phase synchronicity of the auditory steady-state response is extracted, or long-latency event-related potentials (ERPs) are used to assess the brain's response to auditory stimuli. However, in practical applications, these different monitoring modes each have their own emphasis on real-time performance and information richness. For EEP monitoring, which contains rich cognitive information, to avoid aliasing of potential waveforms evoked by adjacent stimuli affecting feature extraction, the stimulus presentation interval is usually longer than the duration of the auditory evoked potential. While this long-interval stimulus presentation method can obtain clear waveforms, the effective sample size obtained per unit time is relatively limited, restricting the data update rate and posing a significant risk of detection lag or missed detection in scenarios with rapid changes in anesthetic status. Summary of the Invention

[0004] In view of this, this application provides a method, system, device and medium for monitoring the depth of anesthesia based on auditory response, in order to solve the above problems.

[0005] Firstly, a method for monitoring the depth of anesthesia based on auditory response is provided, the method comprising:

[0006] A complex auditory stimulus sequence was constructed and output to the subject. The complex auditory stimulus sequence consisted of a steady-state evoked segment and a cognitive exploration segment that alternated on the time axis. The cognitive exploration segment was configured as a rapid presentation mode with stimulus presentation intervals shorter than the duration of auditory evoked potentials.

[0007] Continuous scalp electroencephalogram (EEG) signals were collected from subjects under the influence of a complex auditory stimulation sequence.

[0008] The first data segment corresponding to the steady-state evoked segment is extracted from the continuous scalp EEG signal. Steady-state auditory evoked response features are extracted from the first data segment. The first anesthesia depth index is calculated based on the steady-state auditory evoked response features.

[0009] The second data segment corresponding to the cognitive exploration segment is extracted from the continuous scalp EEG signal. The second data segment is processed using an overlapping signal separation algorithm to obtain the event-related potential waveform. Temporal features are extracted from the event-related potential waveform, and the second anesthesia depth index is calculated based on the temporal features.

[0010] A first confidence weight for the first anesthesia depth index is determined based on the signal quality of the first data segment, and a second confidence weight for the second anesthesia depth index is determined based on the residual noise level of the second data segment.

[0011] The first and second anesthesia depth indices are weighted and fused using the first confidence weight and the second confidence weight to generate and output the anesthesia depth index.

[0012] The above technical solution, by alternately presenting steady-state evoked segments and cognitive probing segments within the same complex auditory stimulus sequence, enables subjects to simultaneously generate steady-state auditory evoked responses and event-related potential responses under continuous stimulation. This design can obtain two types of neural activity signals per unit time, thereby reflecting both physiological activity at the brainstem level and cognitive responses at the cortical level without extending the experimental period. By combining confidence weights determined by signal quality and noise levels, the proportion of indicators can be automatically adjusted according to the real-time signal state, ensuring that the generated anesthesia depth index remains smooth and continuous even with signal fluctuations, reducing misjudgments and delays in real-time monitoring.

[0013] Optionally, construct a complex auditory stimulus sequence, specifically including:

[0014] Generate a basic sound signal that oscillates at a preset carrier frequency, use a preset modulation frequency to perform amplitude modulation on the basic sound signal to obtain a steady-state sound signal, and use the steady-state sound signal as a steady-state induction segment;

[0015] Based on a preset random probability model, a random sequence consisting of standard and deviated stimulus sounds is generated, and the random sequence is mapped to a fast short sound sequence with non-periodic shaking at stimulus presentation intervals within a preset random range.

[0016] Rapid short sound sequences are used as cognitive exploration segments, and steady-state evoked segments and cognitive exploration segments are alternately spliced ​​together according to a preset time ratio to construct a composite auditory stimulus sequence.

[0017] The above technical solution, by designing the steady-state evoked segment as an amplitude-modulated signal with a fixed carrier and modulation frequency, can obtain a high signal-to-noise ratio and stable steady-state response component in the frequency domain. Simultaneously, the cognitive exploration segment is designed as a random short sound sequence composed of standard and deviation tones, with the stimulation interval set shorter than the duration of the evoked potential, allowing cognitively relevant potentials to be continuously excited at a high stimulation rate. This sequence construction method increases stimulation density without introducing waveform aliasing, enabling the system to collect more effective samples in a shorter time, thus improving the temporal resolution and data update rate of anesthesia depth monitoring.

[0018] Optionally, the second data segment is processed using an overlapping signal separation algorithm to obtain the event-related potential waveform, specifically including:

[0019] Extract the start time of each rapid short sound in the cognitive exploration segment of the composite auditory stimulus sequence, and construct a Toplitz structural design matrix describing the stimulus time distribution based on the start time;

[0020] Construct a system of linear convolutional equations with the waveform of the correlation potential of the single event to be solved as the unknown variable, the design matrix as the coefficient matrix, and the second data segment as the observation value;

[0021] The loss function is constructed by combining the prediction error of the linear convolution equation system with the Tikhonov regularization term. The optimal solution vector of the unknown variables is obtained by minimizing the loss function. The optimal solution vector is used as the event-related potential waveform after removing the interference of overlapping adjacent stimuli.

[0022] The above technical solution constructs a Toplitz structural design matrix describing the temporal distribution of stimuli during cognitive exploration, modeling overlapping EEG responses as a system of linear convolutional equations. This computationally clarifies the temporal weight relationships corresponding to each stimulus. By combining this with Tikhonov regularization constraints to minimize the loss function, it effectively separates temporally superimposed event-related potential components, eliminating waveform overlap issues under rapid stimuli. The processing result preserves the true response waveform corresponding to a single stimulus, resulting in higher signal-to-noise ratio and higher temporal localization accuracy for subsequently extracted temporal features.

[0023] Optionally, time-domain features are extracted from the event-related potential waveform, and a second anesthesia depth index is calculated based on the time-domain features, specifically including:

[0024] The event-related potential waveforms were classified, superimposed, and averaged to obtain the standard response waveforms corresponding to the standard stimulus sounds in the cognitive exploration segment and the deviation response waveforms corresponding to the deviation stimulus sounds in the cognitive exploration segment.

[0025] The difference between the deviation response waveform and the standard response waveform is calculated to obtain the mismatch negative wave;

[0026] Search for the trough amplitude of the mismatched negative wave within a preset latency window, and calculate the area under the curve of the mismatched negative wave within a preset time period.

[0027] The amplitude of the trough and the area under the curve are input into a preset support vector machine regression model to obtain a second anesthetic depth index that reflects the degree of cortical cognitive inhibition.

[0028] The above technical solution, by classifying and averaging the separated event-related potentials according to standard and deviation tones, can obtain the average response waveforms of the two sound stimuli. The mismatch negative wave obtained after calculating the difference between the two can reflect the auditory system's recognition response to changes in stimulus. Further extraction of the trough amplitude and area under the curve as quantitative features, and inputting them into a support vector machine regression model for mapping, can transform these time-domain signal features into continuous anesthesia depth indicators, thereby realizing the quantitative calculation of the intensity of cortical cognitive response and reflecting changes in the level of consciousness under anesthesia.

[0029] Optionally, a first confidence weight for the first anesthesia depth index is determined based on the signal quality of the first data segment, and a second confidence weight for the second anesthesia depth index is determined based on the residual noise level of the second data segment, specifically including:

[0030] Perform a fast Fourier transform on the first data segment to calculate the ratio of the spectral power density at the preset modulation frequency to the average power density of the adjacent frequency band, and use the ratio as the signal-to-noise ratio index of the first data segment.

[0031] The design matrix is ​​convolved with the event-related potential waveform to obtain reconstructed data. The difference between the second data segment and the reconstructed data is calculated to obtain the residual sequence. The variance of the residual sequence is used as the residual noise level.

[0032] The signal-to-noise ratio index is mapped to a first confidence weight using a preset first curve function, and the residual noise level is mapped to a second confidence weight using a preset second curve function.

[0033] The above technical solution calculates the signal-to-noise ratio at the modulation frequency using a Fast Fourier Transform (FFT) on the steady-state signal, thus obtaining the energy concentration of the steady-state induced response. Simultaneously, it reconstructs the event-related potential waveform via convolution and compares it with the original signal, calculating the residual variance to reflect the noise intensity in the cognitive segment data that is not explained by the model. By mapping these two indicators to confidence weights, the system can automatically allocate weights based on the actual quality of each signal segment, ensuring that indicators with high signal reliability account for a larger proportion in the fusion process, thereby guaranteeing the stability and accuracy of the output results at different stages.

[0034] Optionally, the difference between the second data segment and the reconstructed data is calculated to obtain a residual sequence, and the variance of the residual sequence is used as the residual noise level. Specifically, this includes:

[0035] Calculate the amplitude difference between the second data segment and the reconstructed data at each sampling point to generate the original residual sequence;

[0036] The original residual sequence is traversed using a sliding window algorithm to identify and remove abnormal data points whose amplitude exceeds the preset artifact threshold, thus obtaining the cleaned residual sequence.

[0037] Calculate the statistical variance of the remaining data points in the cleaned residual sequence, and determine the statistical variance as the residual noise level.

[0038] The above technical solution obtains a residual sequence by calculating the point-by-point difference between the second data segment and the reconstructed data, and uses a sliding window algorithm to remove outliers with amplitudes exceeding the artifact threshold, effectively removing artifacts caused by motion, poor electrode contact, or transient interference. Calculating the variance of the cleaned residual sequence yields a quantitative indicator reflecting the actual noise level, thus reducing the corresponding confidence weight when noise is high. This method makes the noise assessment results closer to the true signal state, ensuring a more reasonable weight allocation for different data segments during fusion calculations.

[0039] Optionally, the first and second anesthesia depth indices are weighted and fused using a first confidence weight and a second confidence weight to generate and output an anesthesia depth index, specifically including:

[0040] The first weighted value is obtained by weighting the first anesthesia depth index with the first confidence weight, and the second weighted value is obtained by weighting the second anesthesia depth index with the second confidence weight. The sum of the first weighted value and the second weighted value is calculated to obtain the instantaneous fusion index.

[0041] Construct a Kalman filter that includes state prediction equations and measurement update equations;

[0042] Calculate the sum of the first confidence weight and the second confidence weight, and adjust the measurement noise covariance matrix in the measurement update equation based on the sum;

[0043] Calculate the prior state estimate using the state prediction equation;

[0044] The anesthesia depth index is obtained by using the measurement update equation, combined with the prior state estimate, the instantaneous fusion index, and the adjusted measurement noise covariance matrix.

[0045] The above technical solution involves weighting and summing two types of anesthesia depth indices with their respective confidence weights and inputting the sum into a Kalman filter. The filter's state prediction and measurement update equations allow for smooth fusion of results over time. By dynamically adjusting the measurement noise covariance matrix based on the total confidence score, the system automatically reduces the update amplitude when signal quality is low, thus avoiding abrupt changes in output values. This method ensures that the final anesthesia depth index is continuous over time, stable under signal interference, and can reflect subtle changes in anesthesia depth in real time.

[0046] Secondly, an anesthesia depth monitoring system based on auditory response is provided, the system comprising:

[0047] The stimulus output module is configured to construct a complex auditory stimulus sequence and output the complex auditory stimulus sequence to the subject. The complex auditory stimulus sequence consists of a steady-state evoked segment and a cognitive exploration segment that alternate on the time axis. The cognitive exploration segment is configured to be a rapid presentation mode with stimulus presentation intervals shorter than the duration of the auditory evoked potential.

[0048] The EEG acquisition module is configured to acquire continuous scalp EEG signals of the subject under the action of a compound auditory stimulation sequence;

[0049] The steady-state analysis module is configured to extract a first data segment corresponding to the steady-state evoked segment from continuous scalp EEG signals, extract steady-state auditory evoked response features from the first data segment, and calculate a first anesthesia depth index based on the steady-state auditory evoked response features.

[0050] The cognitive analysis module is configured to extract a second data segment corresponding to the cognitive exploration segment from continuous scalp EEG signals, process the second data segment using an overlapping signal separation algorithm to obtain event-related potential waveforms, extract time-domain features from the event-related potential waveforms, and calculate a second anesthesia depth index based on the time-domain features.

[0051] The confidence assessment module is configured to determine a first confidence weight for a first anesthesia depth index based on the signal quality of a first data segment, and to determine a second confidence weight for a second anesthesia depth index based on the residual noise level of a second data segment.

[0052] The fusion output module is configured to use a first confidence weight and a second confidence weight to perform weighted fusion of the first anesthesia depth index and the second anesthesia depth index, and generate and output the anesthesia depth index.

[0053] Thirdly, an electronic device is provided, including a processor, a memory, a user interface, and a network interface, wherein the memory is used to store instructions, the user interface and the network interface are both used to communicate with other devices, and the processor is used to execute the instructions stored in the memory to cause the electronic device to perform the method as described in any of the above.

[0054] Fourthly, a computer-readable storage medium is provided, the computer-readable storage medium storing instructions that, when executed, perform the method as described in any of the preceding claims.

[0055] In summary, implementing one or more technical solutions provided in this application has at least the following technical effects or advantages:

[0056] By employing a composite stimulation design and a dual-channel signal processing flow, the system can simultaneously acquire brain functional information at different levels with high sampling density under continuous stimulation, avoiding data discontinuity issues caused by mode switching in traditional methods. The algorithm implementation balances frequency domain feature extraction and time domain waveform separation, enabling stable recovery of key response components related to anesthesia depth in noisy environments and reducing the sensitivity of signal acquisition to external interference. Furthermore, through a dynamic fusion mechanism of confidence weighting and Kalman filtering, the output anesthesia depth index remains smooth, continuous, and up-to-date in real time. This improves the temporal resolution, signal stability, and information integrity of anesthesia depth monitoring without increasing the number of electrodes or experimental time. It provides clinicians with an objective anesthesia assessment method that is real-time, noise-resistant, and cognitively sensitive, and can be extended to surgical anesthesia management, consciousness impairment assessment, and neurological function monitoring. Attached Figure Description

[0057] Figure 1 This is an exemplary system architecture diagram of an anesthesia depth monitoring method or an anesthesia depth monitoring system based on auditory response, which applies the present application.

[0058] Figure 2 This is a flowchart of an anesthesia depth monitoring method based on auditory response disclosed in this application;

[0059] Figure 3 This is a schematic diagram of a module of an anesthesia depth monitoring system based on auditory response disclosed in this application;

[0060] Figure 4 This is a schematic diagram of the structure of an electronic device disclosed in this application.

[0061] Figure reference numerals: 100, System architecture; 101, First terminal device; 102, Second terminal device; 103, Third terminal device; 104, Network; 105, Server; 301, Stimulation output module; 302, EEG acquisition module; 303, Steady-state analysis module; 304, Cognitive analysis module; 305, Confidence assessment module; 306, Fusion output module; 401, Processor; 402, Communication bus; 403, User interface; 404, Network interface; 405, Memory. Detailed Implementation

[0062] To enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments.

[0063] In the description of the embodiments of this application, the words "for example" or "for instance" are used to indicate examples, illustrations, or explanations. Any embodiment or design that is described as "for example" or "for instance" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design options. Rather, the use of the words "for example" or "for instance" is intended to present the relevant concepts in a specific manner.

[0064] In the description of the embodiments of this application, the term "multiple" means two or more. For example, multiple systems means two or more systems, and multiple screen terminals means two or more screen terminals. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the indicated technical features. Thus, a feature defined with "first" or "second" may explicitly or implicitly include one or more of that feature. The terms "comprising," "including," "having," and variations thereof all mean "including but not limited to," unless otherwise specifically emphasized.

[0065] Figure 1 An exemplary system architecture diagram is shown, illustrating an embodiment of an anesthesia depth monitoring method or an anesthesia depth monitoring system based on auditory response to which this application can be applied.

[0066] like Figure 1 As shown, the system architecture 100 may include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104, and a server 105. The network 104 is used as a medium to provide communication links between the terminal devices 101, 102, 103, and the server 105. The network 104 may include various connection types, such as wired or wireless communication links or fiber optic cables, etc.

[0067] Users can use terminal devices 101, 102, and 103 to interact with server 105 via network 104 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 101, 102, and 103, such as model training applications, video recognition applications, web browser applications, social platform software, etc.

[0068] Terminal devices 101, 102, and 103 can be either hardware or software. When terminal devices 101, 102, and 103 are hardware, they can be various electronic devices with displays, including but not limited to smartphones, tablets, e-book readers, MP3 (Moving Picture Experts Group Audio Layer III) players, MP4 (Moving Picture Experts Group Audio Layer IV) players, laptops, and desktop computers, etc. When terminal devices 101, 102, and 103 are software, they can be installed in the aforementioned electronic devices. They can be implemented as multiple software programs or software modules (e.g., multiple software programs or software modules used to provide distributed services) or as a single software program or software module. No specific limitations are imposed here.

[0069] When terminals 101, 102, and 103 are hardware devices, video capture devices can also be installed on them. These video capture devices can be various devices capable of capturing video, such as cameras, sensors, etc. Users can use the video capture devices on terminals 101, 102, and 103 to capture video.

[0070] Server 105 can be a server that provides various services, such as a backend server for processing data displayed on terminal devices 101, 102, and 103. The backend server can analyze and process the received data and can feed back the processing results (such as recognition results) to the terminal devices.

[0071] It should be noted that a server can be either hardware or software. When the server is hardware, it can be implemented as a distributed server cluster consisting of multiple servers, or as a single server. When the server is software, it can be implemented as multiple software programs or software modules (e.g., multiple software programs or software modules used to provide distributed services), or as a single software program or software module. No specific limitations are made here.

[0072] It should be understood that Figure 1The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included. In particular, if the target data does not need to be obtained remotely, the above system architecture may exclude the network and include only terminal devices or servers.

[0073] Figure 2 This is a flowchart illustrating an anesthesia depth monitoring method based on auditory response, as described in this application. This method can be implemented using a computer program, a microcontroller, or run on an anesthesia depth monitoring system based on auditory response. The computer program can be integrated into an application or run as a standalone utility application. The specific steps of an anesthesia depth monitoring method based on auditory response are described in detail below.

[0074] S201: Construct a complex auditory stimulus sequence and output the complex auditory stimulus sequence to the subject. The complex auditory stimulus sequence consists of a steady-state evoked segment and a cognitive exploration segment that alternate on the time axis. The cognitive exploration segment is configured as a rapid presentation mode with stimulus presentation intervals shorter than the duration of auditory evoked potentials.

[0075] For example, time-domain splicing technology is used to integrate periodic modulated signals used to induce primary neural responses with fast random pulse sequences used to test higher cognitive abilities into a single audio stream, wherein the pulse sequence is presented in a high-frequency non-isochronous manner to form a dense exploration of the auditory pathway, thereby constructing a composite stimulus source that can simultaneously stimulate different levels of EEG features.

[0076] In one possible implementation, constructing a composite auditory stimulus sequence specifically includes: generating a basic sound signal that oscillates at a preset carrier frequency; using a preset modulation frequency to amplitude modulate the basic sound signal to obtain a steady-state sound signal, and using the steady-state sound signal as a steady-state evoked segment; generating a random sequence composed of standard stimulus sounds and deviated stimulus sounds based on a preset random probability model, and mapping the random sequence to a rapid short sound sequence with a stimulus presentation interval that jitters non-periodically within a preset random range; using the rapid short sound sequence as a cognitive exploration segment, and alternately splicing the steady-state evoked segment and the cognitive exploration segment according to a preset time ratio to construct a composite auditory stimulus sequence.

[0077] In this embodiment, the composite auditory stimulation sequence refers to an audio arrangement format that integrates a steady-state evoked paradigm targeting the brainstem and primary auditory cortex with an event-related potential paradigm targeting higher cognitive functions in the time dimension. For example, from the perspective of the physical properties of the signal waveform, this composite auditory stimulation sequence is a composite acoustic excitation waveform composed of a continuous amplitude-modulated signal with periodic envelope oscillation characteristics (corresponding to the steady-state evoked segment) and a discrete transient pulse sequence following a random non-isochronous distribution (corresponding to the cognitive exploration segment), time-division multiplexed and alternately spliced ​​along the time axis. This aims to simultaneously activate and monitor the neural responses at different levels of the subject in a single test.

[0078] Specifically, for steady-state waveform synthesis, the system selects a sine wave of a specific frequency (e.g., 1000Hz) as the basic sound signal, i.e., the carrier wave. Then, a specific modulation frequency (usually 40Hz or a frequency that induces a steady-state auditory response) is applied to the carrier wave amplitude. An amplitude modulation algorithm is used to induce periodic oscillations in the sound wave envelope, thereby generating a steady-state sound signal and defining it as the steady-state evoked segment. Next, the logical arrangement of the cognitive exploration components is performed. Based on the classic Oddball paradigm, a random probability model is set to generate a random sequence consisting of standard stimulus sounds with high probability and deviated stimulus sounds with low probability. To enhance the transient stimulation effect on the brain and prevent adaptation, when the system maps this logical sequence to a physical acoustic signal, it forces the stimulus presentation interval to undergo non-periodic jitter within a set random time window (e.g., between 100ms and 300ms). The resulting irregularly rhythmic, rapid, short sound sequence is defined as the cognitive exploration segment. Based on the preset time ratio required by the experimental design (e.g., the steady-state segment lasts for 30 seconds and the exploration segment lasts for 30 seconds), the generated steady-state evoked segment and cognitive exploration segment are alternately spliced ​​and cyclically assembled on the time axis to construct the final composite auditory stimulus sequence to be played to the subjects.

[0079] S202: Collect continuous scalp EEG signals from the subject under the influence of a complex auditory stimulation sequence.

[0080] In this embodiment, continuous scalp EEG signal refers to a data stream of potential changes recorded by non-invasive electrodes attached to the scalp surface, reflecting the synchronous firing activity of a group of neurons in the brain. For example, this signal can be characterized as raw time-series voltage values ​​that have not been segmented, filtered, or averaged, encompassing evoked potential components highly correlated with external auditory stimuli as well as the brain's own spontaneous background rhythms.

[0081] Specifically, to capture the brain's real-time response to auditory stimuli, EEG acquisition electrodes need to be pre-positioned at specific locations on the subject's scalp (such as key recording points in the frontal region Fz, central region Cz, or parietal region Pz) and reference locations (such as the mastoid process), following the international 10-20 system or other standard lead configurations. When the aforementioned constructed composite auditory stimulation sequence is continuously played to the subject through headphones or speakers, the subject's auditory pathways and related cortex will produce corresponding neurophysiological responses under the continuous alternation of different sound components (i.e., steady-state evoked segments and cognitive exploration segments) in the sequence. At this time, the hardware acquisition interface is activated to amplify the weak analog bioelectrical signals captured by the electrodes with low noise, perform hardware filtering, and perform analog-to-digital conversion (A / D conversion). The analog signals are then converted into digital form at a preset high sampling frequency (e.g., 1000Hz or higher), thereby recording the continuous scalp EEG signals throughout the entire stimulation process in real time, completely, and uninterruptedly.

[0082] S203: Extract the first data segment corresponding to the steady-state evoked segment from the continuous scalp EEG signal, extract the steady-state auditory evoked response features from the first data segment, and calculate the first anesthesia depth index based on the steady-state auditory evoked response features.

[0083] In the embodiments of this application, steady-state auditory evoked response characteristics refer to frequency domain or time-frequency domain parameters that reflect the ability of a group of brain neurons to produce a phase-locked following response to periodic auditory stimuli. For example, these characteristics are typically manifested as the peak energy spectral density, phase coherence value, or signal-to-noise ratio at a specific frequency of the electroencephalogram (EEG) at the stimulation modulation frequency (e.g., 40 Hz), used to characterize the excitability state of the primary auditory cortex and brainstem pathways.

[0084] Specifically, based on the timestamps or trigger markers during the playback of the complex auditory stimulus sequence, the acquired continuous scalp EEG signals are temporally localized and segmented. EEG data that completely correspond to the steady-state evoked response segment on the time axis are identified and extracted, and this portion of data is defined as the first data segment. This first data segment undergoes signal preprocessing and frequency domain analysis (such as Fast Fourier Transform), focusing on detecting and separating the steady-state component synchronized with the modulation frequency of the stimulus carrier, thereby quantifying and extracting the steady-state auditory evoked response characteristics.

[0085] In one specific embodiment of this application, when calculating the first anesthesia depth index, a nonlinear mapping model based on the modified Sigmoid function is used to transform frequency domain features into a quantitative score that conforms to clinical practice. The specific calculation process is as follows: 1. Feature quantization: First, feature quantization is performed to calculate the signal power P at a preset modulation frequency (e.g., 40Hz) in the spectrum of the first data segment. signal The average noise power P of its adjacent frequency band (e.g., within the 38Hz-42Hz range, excluding the 40Hz point)noise Next, the signal-to-noise ratio is calculated using the formula. Calculate the logarithmic signal-to-noise ratio (SNR). Finally, perform index mapping. To map the unbounded SNR to a quantization index ranging from 0 to 100, the following formula is used. The first anesthesia depth index, Idx1, is calculated. Here, α is the slope parameter (preferably 0.5 in this embodiment), used to control the sensitivity of the index to changes in signal-to-noise ratio; β is the threshold parameter (preferably 6 dB in this embodiment), representing the index at an intermediate level (i.e., 50 points) when the signal-to-noise ratio reaches this value, used to calibrate the critical point between wakefulness and anesthesia. Through this model, the system can standardize physiological responses of different intensities into a uniform anesthesia depth value.

[0086] S204: Extract the second data segment corresponding to the cognitive exploration segment from the continuous scalp EEG signal, process the second data segment using the overlapping signal separation algorithm to obtain the event-related potential waveform, extract the time-domain features from the event-related potential waveform, and calculate the second anesthesia depth index based on the time-domain features.

[0087] For example, to address the signal overlap problem caused by rapid auditory stimulation, a dealiasing technique based on linear deconvolution is used to restore the pure single evoked response from continuous EEG. Furthermore, the brain's automatic discrimination response to novel stimuli is captured to extract mismatch negative wave features. Finally, the electrophysiological feature is transformed into a quantitative indicator representing the level of cortical cognitive processing through a regression model.

[0088] In one possible implementation, the second data segment is processed using an overlap signal separation algorithm to obtain the event-related potential waveform. Specifically, this includes: extracting the start time of each rapid short sound within the cognitive exploration segment of the complex auditory stimulus sequence; constructing a Toplitz structure design matrix describing the stimulus time distribution based on the start time; constructing a system of linear convolution equations with the single event-related potential waveform to be solved as the unknown variable, the design matrix as the coefficient matrix, and the second data segment as the observation; constructing a loss function based on the prediction error of the linear convolution equations and the Tikhonov regularization term; calculating the optimal solution vector of the unknown variable by minimizing the loss function; and using the optimal solution vector as the event-related potential waveform after removing the interference of overlapping adjacent stimuli.

[0089] In this embodiment, the Toplitz structural design matrix refers to a special mathematical matrix with constant diagonal elements, used here to simulate the convolution operation process in a time-invariant system using linear algebra methods. For example, this matrix can be understood as a two-dimensional array composed of pulse sequences describing the timing of stimulus occurrences after multiple time shifts and rearrangements, where each column represents the potential delay effect of stimuli presented at different times on subsequent EEG signals, thereby transforming complex temporal overlap relationships of signals into a standard system of linear equations.

[0090] Specifically, because the cognitive exploration segment employs a rapid and non-periodic stimulus presentation method, the brain responses evoked by adjacent stimuli become aliased in time, making it impossible to obtain a pure waveform through simple superposition and averaging. Therefore, it is necessary to accurately pinpoint the start time (i.e., time-stamps) of each rapid short sound within the cognitive exploration segment of the complex auditory stimulus sequence. Using this time-point information, a Toeplitz structure design matrix with Toeplitz properties is constructed. This matrix mathematically represents the distribution and lag relationships of stimulus events along the time axis. Based on the linear system assumption (i.e., the total response equals the linear superposition of individual responses), a system of linear convolution equations is constructed. The second data segment (i.e., the acquired aliased EEG signals) serves as the known observation vector, the design matrix as the coefficient matrix, and the single-event-related potential waveform to be solved as the unknown variable vector. To address the ill-posedness in this deconvolution problem and suppress noise amplification, direct inversion is not possible. Instead, a loss function is constructed that includes prediction error (model fit) and Tikhonov regularization (smoothness constraint of the solution). The function is solved by mathematical optimization algorithm to find the optimal solution vector that minimizes the loss function. This vector is the event-related potential waveform that is mathematically deconstructed and has removed the interference of overlapping adjacent stimuli.

[0091] In one specific embodiment of this application, the process of using the Toplitz matrix to solve the linear convolution equations to obtain a pure event-related potential waveform involves the following detailed mathematical construction and solution steps: First, a design matrix is ​​constructed, assuming that the number of sampling points in the second data segment is N, the length of the single event-related potential waveform is L (corresponding to the duration of the evoked response), and the total number of stimulus events is K; a Toplitz design matrix X with dimension N×L is constructed, and the elements X in the matrix... i，j Defined as: if at time t i If a stimulus event occurred at one of the previous j sampling points, then X i，j =1, otherwise X i，j =0, this matrix not only encodes the start time of the stimulus, but also implicitly contains information about different delay times through column shifts. Next, the system of equations is defined to establish a linear model. , where y is the N×1 observation vector (i.e., the second data segment), β is the L×1 unknown vector to be solved (i.e., the single-event correlated potential waveform), and ϵ is the residual noise. Finally, the loss function is constructed and solved, resulting in a loss function that includes the Tikhonov regularization term. Where λ is the regularization parameter used to control the smoothness of the solution to prevent overfitting, and D is the second-order difference matrix or identity matrix; by taking the derivative with respect to β and setting it to zero, the analytical solution (closed-form solution) of this convex optimization problem is obtained as follows: The system directly calculates this formula through matrix operations, thus enabling a one-time calculation of the event-related potential waveform after removing the linear superposition effect of adjacent responses. This effectively solves the problem of waveform overlap under rapid stimulation.

[0092] In one possible implementation, temporal features are extracted from the event-related potential waveforms, and a second anesthesia depth index is calculated based on these temporal features. Specifically, this includes: classifying and averaging the event-related potential waveforms to obtain a standard response waveform corresponding to a standard stimulus sound in the cognitive exploration segment and a deviation response waveform corresponding to a deviation stimulus sound in the cognitive exploration segment; calculating the difference between the deviation response waveform and the standard response waveform to obtain a mismatch negative wave; searching for the trough amplitude of the mismatch negative wave within a preset latency window and calculating the area under the curve of the mismatch negative wave within a preset time period; inputting the trough amplitude and area under the curve into a preset support vector machine regression model to map and obtain a second anesthesia depth index reflecting the degree of cortical cognitive inhibition.

[0093] In the embodiments of this application, mismatch negativity (MMN) refers to an endogenous event-related potential component that reflects the automatic detection and pre-attentional processing of changes in environmental acoustic rules by the auditory cortex under unconscious involvement. For example, this waveform typically manifests as a negative voltage deflection within a latency period of approximately 100ms to 250ms after stimulation, following the difference between the EEG response induced by a low-probability deviation stimulus and the EEG response induced by a high-probability standard stimulus in an auditory oddball paradigm (such as the weird ball paradigm).

[0094] Specifically, to accurately assess the retention of higher cognitive functions in the brain, the event-related potential waveforms of each isolated single event need to be classified and aggregated based on the event labels of the stimulus sequence in the cognitive exploration segment. Multiple waveforms corresponding to frequently recurring background sounds (i.e., standard stimuli) and multiple waveforms corresponding to occasionally occurring variant sounds (i.e., deviation stimuli) are then superimposed in the time domain and averaged. This process suppresses random background noise while obtaining standard and deviation response waveforms with high signal-to-noise ratios. Subsequently, a difference operation is performed, subtracting the voltage amplitude of the standard response waveform at the corresponding moment from the voltage amplitude of the deviation response waveform to generate a difference waveform, or MMN, that characterizes the brain's ability to automatically distinguish novel information. Within the typical physiological occurrence range of the MMN (i.e., the preset latency window), the negative extreme points of this difference waveform are automatically searched to determine the trough amplitude. Simultaneously, the absolute value or envelope of the waveform's amplitude over a preset time period is integrated to obtain the area under the curve. These two quantitative features are input into a pre-trained support vector machine (SVM) regression model that establishes a non-linear mapping relationship between electrophysiological features and levels of consciousness. The model then outputs a numerical result that quantitatively reflects the degree of cortical cognitive inhibition in the subject, i.e., the second anesthesia depth index. It should be noted that the SVM regression model is pre-trained on a standard anesthesia dataset. This dataset contains auditory evoked response features (as input) of multiple subjects throughout the entire process of awakening, induction, maintenance, and recovery, along with corresponding gold-standard anesthesia depth indices (such as BIS values ​​or blood drug concentrations, as labels). The model is obtained through non-linear regression fitting using radial basis function kernels.

[0095] S205: Determine a first confidence weight for the first anesthesia depth index based on the signal quality of the first data segment, and determine a second confidence weight for the second anesthesia depth index based on the residual noise level of the second data segment.

[0096] For example, the quality of the collected EEG data is quantified from two dimensions: frequency domain features and time domain error. On the one hand, the signal-to-noise ratio is evaluated based on the energy concentration in a specific frequency band. On the other hand, the background noise level is evaluated by calculating the degree of deviation between the observed data and the theoretical model. Then, the two quality indicators are mapped to corresponding weight coefficients using a preset functional relationship to achieve differentiated rating of the reliability of anesthesia indicators from different sources.

[0097] In one possible implementation, determining a first confidence weight for a first anesthesia depth index based on the signal quality of a first data segment, and determining a second confidence weight for a second anesthesia depth index based on the residual noise level of a second data segment, specifically includes: performing a fast Fourier transform on the first data segment, calculating the ratio of the spectral power density at a preset modulation frequency to the average power density of adjacent frequency bands, and using the ratio as the signal-to-noise ratio index of the first data segment; performing a convolution operation on the design matrix and the event-related potential waveform to obtain reconstructed data, calculating the difference between the second data segment and the reconstructed data to obtain a residual sequence, and statistically analyzing the variance of the residual sequence as the residual noise level; mapping the signal-to-noise ratio index to a first confidence weight using a preset first curve function, and mapping the residual noise level to a second confidence weight using a preset second curve function.

[0098] In this embodiment, the confidence weight refers to a numerical coefficient used to quantify the reliability of a specific measurement result or calculated index in the multi-source information fusion decision-making process. For example, this weight is typically a normalized value between 0 and 1. When the signal-to-noise ratio of a certain EEG signal is extremely high or the model fitting error is extremely small, the system assigns a higher weight value to the corresponding anesthesia depth index, indicating that the index more accurately reflects the subject's brain state; conversely, it reduces the weight of the index in the final score.

[0099] Specifically, to ensure the robustness of the final evaluation results, the influence of each indicator needs to be dynamically adjusted based on the purity of the signal itself. For frequency domain characteristics, a Fast Fourier Transform is performed on the first data segment to convert the time-domain waveform into a spectrum, accurately locating the preset modulation frequency (i.e., the carrier frequency of the auditory stimulus). The ratio of the spectral power density at this frequency point to the average power density of its two adjacent frequency bands is calculated. A higher ratio indicates a more significant evoked response, and this is identified as the signal-to-noise ratio (SNR) indicator for the first data segment. For time domain characteristics, to evaluate the goodness of fit of the de-overlap algorithm, the design matrix describing the stimulus timing is re-convolved with the extracted pure event-related potential waveform to generate reconstructed data under ideal conditions (i.e., the theoretically predicted signal). The point-by-point difference between the original acquired second data segment and this reconstructed data is calculated to obtain the residual sequence, and the variance of this sequence is statistically analyzed. A larger variance indicates more clutter that the model fails to explain, and this is used as the residual noise level. The quality parameters calculated above are input into the mapping module. The signal-to-noise ratio index is converted into a first confidence weight for the first index using a preset first curve function (usually a positive correlation mapping, such as a sigmoid normalization function). The residual noise level is converted into a second confidence weight for the second index using a preset second curve function (usually a negative correlation mapping, such as a negative exponential decay function).

[0100] In one possible implementation, the difference between the second data segment and the reconstructed data is calculated to obtain a residual sequence, and the variance of the residual sequence is used as the residual noise level. Specifically, this includes: calculating the amplitude difference between the second data segment and the reconstructed data at each sampling point to generate an original residual sequence; using a sliding window algorithm to traverse the original residual sequence, identifying and removing abnormal data points whose amplitude exceeds a preset artifact threshold to obtain a cleaned residual sequence; calculating the statistical variance of the remaining data points in the cleaned residual sequence, and determining the statistical variance as the residual noise level.

[0101] In this embodiment, the residual sequence refers to a time series obtained through mathematical difference operations, used to represent the remaining components in the original acquired signal that could not be explained or fitted by the theoretical model. For example, this sequence typically includes background EEG activity, environmental electromagnetic interference, and non-physiological artifact noise, and its amplitude fluctuations directly reflect the background noise intensity of the current data acquisition environment, i.e., the degree of inconsistency between the actual observed value and the theoretical predicted value.

[0102] Specifically, to accurately quantify the background noise intensity after removing event-related potentials, signal subtraction is required. This involves calculating the amplitude difference between the actual acquired second data segment and the model-generated reconstructed data at each sampling time, thereby eliminating induced response components from the signal and generating an original residual sequence containing all potential noise components. To avoid sudden, large-amplitude interference (such as abrupt changes caused by electrooculogram artifacts or poor electrode contact) distorting the overall background noise assessment, a sliding window algorithm is used to traverse the original residual sequence segment by segment along the time axis. This automatically detects data features within the window, accurately identifies and removes outlier data points whose absolute amplitude exceeds a preset artifact threshold, thus obtaining a cleaned residual sequence free of high-amplitude artifact interference. Statistical analysis is then performed on all remaining data points in this sequence, calculating their statistical variance to measure the data dispersion. The calculated statistical variance value is directly determined as the residual noise level.

[0103] S206: The first and second anesthesia depth indices are weighted and fused using the first confidence weight and the second confidence weight to generate and output the anesthesia depth index.

[0104] For example, based on the differences in signal purity of each data segment, the fusion ratio of the corresponding anesthesia depth index is adaptively adjusted to obtain a preliminary comprehensive observation value. A filtering algorithm is then introduced to smooth and correct the observation value in the time domain according to the overall confidence level, thereby outputting a final evaluation result with stronger anti-interference capabilities.

[0105] In one possible implementation, a weighted fusion of a first anesthesia depth index and a second anesthesia depth index is performed using a first confidence weight and a second confidence weight to generate and output an anesthesia depth index. Specifically, this includes: weighting the first anesthesia depth index using the first confidence weight to obtain a first weighted value; weighting the second anesthesia depth index using the second confidence weight to obtain a second weighted value; calculating the sum of the first and second weighted values ​​to obtain an instantaneous fusion index; constructing a Kalman filter containing a state prediction equation and a measurement update equation; calculating the sum of the first and second confidence weights, and adjusting the measurement noise covariance matrix in the measurement update equation based on the sum; calculating a priori state estimates using the state prediction equation; and calculating the anesthesia depth index using the measurement update equation, combining the priori state estimates, the instantaneous fusion index, and the adjusted measurement noise covariance matrix.

[0106] In this embodiment, the Kalman filter refers to a recursive algorithm that uses the state equation of a linear system to optimally estimate the system state based on the system's input and output observation data. For example, in this scenario, the filter is constructed as a mathematical model that does not rely solely on the current single measurement value. Instead, it fuses the previous estimate with the current measurement data, and, considering statistical noise and uncertainty in the data, predicts and corrects the final output, thereby reconstructing a smooth and realistic trajectory of anesthesia depth changes from a data stream containing interference.

[0107] Specifically, to achieve accurate fusion of multi-source monitoring data and filter out random fluctuations, a weighted averaging operation is first performed. A first weighted value is obtained by multiplying a first confidence weight (reflecting the quality of the first signal) with a first anesthesia depth index. Simultaneously, a second weighted value is obtained by multiplying a second confidence weight with a second anesthesia depth index. These two weighted values ​​are then summed to obtain an instantaneous fusion index reflecting the comprehensive observation results at the current moment. Following this, a filtering stage is initiated, initializing and constructing a Kalman filter containing a state prediction equation (used to infer the current state based on the previous state) and a measurement update equation (used to correct the inferred state based on the observed values). During this process, the current measurement confidence level needs to be dynamically evaluated, i.e., the sum of the first and second confidence weights is calculated. A higher sum indicates good overall signal quality, while a lower sum indicates significant interference. Based on this sum, the measurement noise covariance matrix in the filter parameters is adjusted inversely (e.g., a larger sum results in a smaller measurement noise parameter, indicating an increase in the weight of the current observation in the posterior estimation). The state prediction equation is used to calculate the prior state estimate for the current moment based on the state at the previous moment. By using the measurement update equation, the weighted correction calculation is performed by comprehensively considering the prior state estimate, the instantaneous fusion index as the actual observation input, and the dynamically adjusted measurement noise covariance matrix, so as to obtain the optimal posterior estimate, which is the final anesthesia depth index.

[0108] In one specific embodiment of this application, the construction and dynamic adjustment mechanism of the Kalman filter used for fusing weighted indices is implemented through the following mathematical model: First, define the state vector. ,in This represents the actual depth of anesthesia at time k. This represents the rate of change in the depth of anesthesia. Next, a priori estimation is performed using the state prediction equation, the formula of which is... as well as ,in Let P be the state transition matrix, Δt be the time step, P be the error covariance matrix, and Q be the process noise covariance matrix. Simultaneously, a dynamic adjustment mechanism for measurement noise is introduced to calculate the total confidence level. (i.e., the sum of the first and second confidence weights), setting the basic measurement noise variance as R0, then the dynamic measurement noise covariance matrix R at the current time is... k Defined as , where ϵ is a small positive number to prevent the denominator from being zero; this formula implements a mechanism for dynamically adjusting the Kalman gain based on signal quality: when the signal quality is good, the weight of the measurement vector in the posterior state estimation is increased; conversely, the weight of the prior state prediction is increased. Finally, the final output is obtained using the measurement update equation, the formula is: as well as , where z kHere, H = [1, 0] represents the instantaneous fusion index, and H is the observation matrix. The final output anesthesia depth index is the value in the updated state vector. This allows for the reconstruction of a smooth and realistic trajectory of anesthesia depth changes within a data stream containing interference.

[0109] Figure 3 This is a schematic diagram of a module of an anesthesia depth monitoring system based on auditory response, as described in an embodiment of this application. This system can be implemented through software, hardware, or a combination of both, forming all or part of the overall system. For example... Figure 3 As shown, the system includes:

[0110] Stimulation output module 301 is configured to construct a complex auditory stimulus sequence and output the complex auditory stimulus sequence to the subject. The complex auditory stimulus sequence consists of a steady-state evoked segment and a cognitive exploration segment that alternate on the time axis. The cognitive exploration segment is configured as a rapid presentation mode with a stimulus presentation interval shorter than the duration of the auditory evoked potential.

[0111] EEG acquisition module 302 is configured to acquire continuous scalp EEG signals of the subject under the action of a compound auditory stimulation sequence;

[0112] The steady-state analysis module 303 is configured to extract a first data segment corresponding to the steady-state evoked segment from the continuous scalp EEG signal, extract steady-state auditory evoked response features from the first data segment, and calculate a first anesthesia depth index based on the steady-state auditory evoked response features.

[0113] The cognitive analysis module 304 is configured to extract a second data segment corresponding to the cognitive exploration segment from the continuous scalp electroencephalogram signal, process the second data segment using an overlapping signal separation algorithm to obtain event-related potential waveforms, extract time-domain features from the event-related potential waveforms, and calculate a second anesthesia depth index based on the time-domain features.

[0114] The confidence assessment module 305 is configured to determine a first confidence weight for a first anesthesia depth index based on the signal quality of a first data segment, and to determine a second confidence weight for a second anesthesia depth index based on the residual noise level of a second data segment.

[0115] The fusion output module 306 is configured to use a first confidence weight and a second confidence weight to perform weighted fusion of the first anesthesia depth index and the second anesthesia depth index, and generate and output the anesthesia depth index.

[0116] Based on the above embodiments, as an optional embodiment, the stimulus output module 301 is specifically used to: generate a basic sound signal that oscillates at a preset carrier frequency, perform amplitude modulation on the basic sound signal using a preset modulation frequency to obtain a steady-state sound signal, and use the steady-state sound signal as a steady-state evoked segment; generate a random sequence composed of standard stimulus sounds and deviated stimulus sounds based on a preset random probability model, and map the random sequence into a rapid short sound sequence with a non-periodic jittering stimulus presentation interval within a preset random range; use the rapid short sound sequence as a cognitive exploration segment, and alternately splice the steady-state evoked segment and the cognitive exploration segment according to a preset time ratio to construct a composite auditory stimulus sequence.

[0117] Based on the above embodiments, as an optional embodiment, the cognitive analysis module 304 is specifically used for: extracting the start time of each rapid short sound in the cognitive exploration segment of the compound auditory stimulus sequence; constructing a Toplitz structure design matrix describing the stimulus time distribution based on the start time; constructing a system of linear convolution equations with the single event-related potential waveform to be solved as the unknown variable, the design matrix as the coefficient matrix, and the second data segment as the observation value; constructing a loss function based on the prediction error of the linear convolution equation system and the Tikhonov regularization term; calculating the optimal solution vector of the unknown variable by minimizing the loss function; and using the optimal solution vector as the event-related potential waveform after removing the interference of overlapping adjacent stimuli.

[0118] Based on the above embodiments, as an optional embodiment, the cognitive analysis module 304 is specifically used to: classify, superimpose, and average the event-related potential waveforms to obtain the standard response waveform corresponding to the standard stimulus sound in the cognitive exploration segment and the deviation response waveform corresponding to the deviation stimulus sound in the cognitive exploration segment; calculate the difference between the deviation response waveform and the standard response waveform to obtain the mismatch negative wave; search for the trough amplitude of the mismatch negative wave within a preset latency window and calculate the area under the curve of the mismatch negative wave within a preset time period; input the trough amplitude and the area under the curve into a preset support vector machine regression model to map and obtain a second anesthesia depth index reflecting the degree of cortical cognitive inhibition.

[0119] Based on the above embodiments, as an optional embodiment, the confidence assessment module 305 is specifically used for: performing a fast Fourier transform on the first data segment, calculating the ratio of the spectral power density at a preset modulation frequency to the average power density of the adjacent frequency band, and using the ratio as the signal-to-noise ratio index of the first data segment; performing a convolution operation on the design matrix and the event-related potential waveform to obtain reconstructed data, calculating the difference between the second data segment and the reconstructed data to obtain a residual sequence, and statistically analyzing the variance of the residual sequence as the residual noise level; mapping the signal-to-noise ratio index to a first confidence weight using a preset first curve function, and mapping the residual noise level to a second confidence weight using a preset second curve function.

[0120] Based on the above embodiments, as an optional embodiment, the confidence assessment module 305 is specifically used to: calculate the amplitude difference between the second data segment and the reconstructed data at each sampling point to generate the original residual sequence; use a sliding window algorithm to traverse the original residual sequence, identify and remove abnormal data points whose amplitude exceeds the preset artifact threshold, and obtain the cleaned residual sequence; calculate the statistical variance of the remaining data points in the cleaned residual sequence, and determine the statistical variance as the residual noise level.

[0121] Based on the above embodiments, as an optional embodiment, the fusion output module 306 is specifically used for: weighting the first anesthesia depth index using a first confidence weight to obtain a first weighted value, weighting the second anesthesia depth index using a second confidence weight to obtain a second weighted value, and calculating the sum of the first weighted value and the second weighted value to obtain an instantaneous fusion index; constructing a Kalman filter that includes a state prediction equation and a measurement update equation; calculating the sum of the first confidence weight and the second confidence weight, and adjusting the measurement noise covariance matrix in the measurement update equation based on the sum; calculating the prior state estimate using the state prediction equation; and calculating the anesthesia depth index using the measurement update equation, combined with the prior state estimate, the instantaneous fusion index, and the adjusted measurement noise covariance matrix.

[0122] It should be noted that the system provided in the above embodiments is only illustrated by the division of the above functional modules. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the system and method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process can be found in the method embodiments, which will not be repeated here.

[0123] This embodiment also discloses an electronic device, as shown in the reference. Figure 4 The electronic device may include: at least one processor 401, at least one communication bus 402, user interface 403, network interface 404, and at least one memory 405.

[0124] The communication bus 402 is used to enable communication between these components.

[0125] The user interface 403 may include a display screen and a camera. Optionally, the user interface 403 may also include a standard wired interface and a wireless interface.

[0126] The network interface 404 may optionally include a standard wired interface or a wireless interface (such as a Wi-Fi interface).

[0127] The processor 401 may include one or more processing cores. The processor 401 connects to various parts of the server using various interfaces and lines, and performs various server functions and processes data by running or executing instructions, programs, code sets, or instruction sets stored in memory 405, and by calling data stored in memory 405. Optionally, the processor 401 may be implemented using at least one hardware form of Digital Signal Processing (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 401 may integrate one or a combination of several of the following: Central Processing Unit (CPU), Graphics Processing Unit (GPU), and modem. The CPU primarily handles the operating system, user interface, and applications; the GPU is responsible for rendering and drawing the content required for display; and the modem handles wireless communication. It is understood that the modem may also be implemented as a separate chip without being integrated into the processor 401.

[0128] The memory 405 may include random access memory (RAM) or read-only memory. Optionally, the memory 405 may include a non-transitory computer-readable storage medium. The memory 405 may be used to store instructions, programs, code, code sets, or instruction sets. The memory 405 may include a program storage area and a data storage area, wherein the program storage area may store instructions for implementing an operating system, instructions for at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above-described method embodiments, etc.; the data storage area may store data involved in the above-described method embodiments, etc. Optionally, the memory 405 may also be at least one storage device located remotely from the aforementioned processor 401. Figure 4 As shown, the memory 405, which serves as a computer storage medium, may include an operating system, a network communication module, a user interface module, and an application program for an anesthesia depth monitoring method based on auditory response.

[0129] exist Figure 4In the electronic device shown, the user interface 403 is mainly used to provide an input interface for the user and to obtain the user input data; while the processor 401 can be used to call an application program stored in the memory 405 for anesthesia depth monitoring based on auditory response. When executed by one or more processors 401, the electronic device performs one or more methods as described in the above embodiments.

[0130] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0131] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0132] In the several embodiments provided in this application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the shown or discussed mutual couplings or direct couplings or communication connections may be through some service interfaces; indirect couplings or communication connections between apparatuses or units may be electrical or other forms.

[0133] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0134] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0135] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory 405 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 of the various embodiments of this application. The aforementioned memory 405 includes various media capable of storing program code, such as a USB flash drive, external hard drive, magnetic disk, or optical disk.

[0136] The foregoing description is merely an exemplary embodiment of this disclosure and should not be construed as limiting the scope of this disclosure. Any equivalent changes and modifications made in accordance with the teachings of this disclosure shall still fall within the scope of this disclosure. Those skilled in the art will readily conceive of other embodiments of this disclosure upon considering the disclosure in this specification. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not described in this disclosure. The specification and embodiments are considered exemplary only, and the scope of this application is defined by the claims.

Claims

1. A method for monitoring the depth of anesthesia based on auditory response, characterized in that, The method includes: A complex auditory stimulus sequence is constructed and output to the subject. The complex auditory stimulus sequence consists of a steady-state evoked segment and a cognitive exploration segment that alternately cycle on the time axis. The cognitive exploration segment is configured as a rapid presentation mode with stimulus presentation intervals shorter than the duration of auditory evoked potentials. Continuous scalp electroencephalogram (EEG) signals were collected from the subject under the influence of the compound auditory stimulation sequence; The first data segment corresponding to the steady-state evoked segment is extracted from the continuous scalp EEG signal, the steady-state auditory evoked response features are extracted from the first data segment, and the first anesthesia depth index is calculated based on the steady-state auditory evoked response features. The second data segment corresponding to the cognitive exploration segment is extracted from the continuous scalp EEG signal. The second data segment is processed using an overlapping signal separation algorithm to obtain event-related potential waveforms. Temporal features are extracted from the event-related potential waveforms, and a second anesthesia depth index is calculated based on the temporal features. A first confidence weight for the first anesthesia depth index is determined based on the signal quality of the first data segment, and a second confidence weight for the second anesthesia depth index is determined based on the residual noise level of the second data segment. The first and second anesthesia depth indices are weighted and fused using the first and second confidence weights to generate and output the anesthesia depth index.

2. The method according to claim 1, characterized in that, The construction of the complex auditory stimulus sequence specifically includes: A basic sound signal oscillates at a preset carrier frequency, and the basic sound signal is amplitude modulated using a preset modulation frequency to obtain a steady-state sound signal, which is then used as the steady-state induction segment. A random sequence consisting of standard and deviated stimulus sounds is generated based on a preset random probability model, and the random sequence is mapped to a fast short sound sequence with non-periodic shaking at stimulus presentation intervals within a preset random range. The rapid short sound sequence is used as the cognitive exploration segment, and the steady-state evoked segment and the cognitive exploration segment are alternately spliced ​​according to a preset time ratio to construct the composite auditory stimulus sequence.

3. The method according to claim 1, characterized in that, The process of using an overlapping signal separation algorithm to process the second data segment to obtain the event-related potential waveform specifically includes: Extract the start time of each rapid short sound in the cognitive exploration segment of the composite auditory stimulus sequence, and construct a Toplitz structural design matrix describing the stimulus time distribution based on the start time; Construct a system of linear convolutional equations with the waveform of the single event-related potential to be solved as the unknown variable, the design matrix as the coefficient matrix, and the second data segment as the observation value; A loss function is constructed based on the prediction error of the linear convolution equation system and the Tikhonov regularization term. The optimal solution vector of the unknown variable is calculated by minimizing the loss function. The optimal solution vector is used as the event-related potential waveform after removing the interference of overlapping adjacent stimuli.

4. The method according to claim 3, characterized in that, The step of extracting time-domain features from the event-related potential waveform and calculating the second anesthesia depth index based on the time-domain features specifically includes: The event-related potential waveforms are classified, superimposed, and averaged to obtain the standard response waveforms corresponding to the standard stimulus sounds in the cognitive exploration segment and the deviation response waveforms corresponding to the deviation stimulus sounds in the cognitive exploration segment, respectively. Calculate the difference between the deviation response waveform and the standard response waveform to obtain the mismatch negative wave; Search for the trough amplitude of the mismatched negative wave within a preset latency window, and calculate the area under the curve of the mismatched negative wave within a preset time period. The trough amplitude and the area under the curve are input into a preset support vector machine regression model to obtain the second anesthesia depth index, which reflects the degree of cortical cognitive inhibition.

5. The method according to claim 3, characterized in that, The step of determining a first confidence weight for the first anesthesia depth index based on the signal quality of the first data segment, and determining a second confidence weight for the second anesthesia depth index based on the residual noise level of the second data segment, specifically includes: Perform a fast Fourier transform on the first data segment to calculate the ratio of the spectral power density at the preset modulation frequency to the average power density of the adjacent frequency band, and use the ratio as the signal-to-noise ratio index of the first data segment. The design matrix is ​​convolved with the event-related potential waveform to obtain reconstructed data. The difference between the second data segment and the reconstructed data is calculated to obtain a residual sequence. The variance of the residual sequence is used as the residual noise level. The signal-to-noise ratio index is mapped to the first confidence weight using a preset first curve function, and the residual noise level is mapped to the second confidence weight using a preset second curve function.

6. The method according to claim 5, characterized in that, The step of calculating the difference between the second data segment and the reconstructed data to obtain a residual sequence, and then calculating the variance of the residual sequence as the residual noise level, specifically includes: Calculate the amplitude difference between the second data segment and the reconstructed data at each sampling point to generate the original residual sequence; The original residual sequence is traversed using a sliding window algorithm to identify and remove abnormal data points whose amplitude exceeds a preset artifact threshold, thus obtaining the cleaned residual sequence. Calculate the statistical variance of the remaining data points in the cleaned residual sequence, and determine the statistical variance as the residual noise level.

7. The method according to claim 1, characterized in that, The step of weighting and fusing the first and second anesthesia depth indices using the first and second confidence weights to generate and output an anesthesia depth index specifically includes: The first weighted value is obtained by weighting the first anesthesia depth index using the first confidence weight, and the second weighted value is obtained by weighting the second anesthesia depth index using the second confidence weight. The sum of the first weighted value and the second weighted value is calculated to obtain the instantaneous fusion index. Construct a Kalman filter that includes state prediction equations and measurement update equations; Calculate the sum of the first confidence weight and the second confidence weight, and adjust the measurement noise covariance matrix in the measurement update equation based on the sum; The prior state estimate is calculated using the state prediction equation. The anesthesia depth index is obtained by using the measurement update equation, combined with the prior state estimate, the instantaneous fusion index, and the adjusted measurement noise covariance matrix.

8. An anesthesia depth monitoring system based on auditory response, characterized in that, The system includes: The stimulus output module is configured to construct a complex auditory stimulus sequence and output the complex auditory stimulus sequence to the subject. The complex auditory stimulus sequence consists of a steady-state evoked segment and a cognitive exploration segment that alternately cycle on the time axis. The cognitive exploration segment is configured to be a rapid presentation mode with a stimulus presentation interval shorter than the duration of the auditory evoked potential. The EEG acquisition module is configured to acquire continuous scalp EEG signals of the subject under the action of the compound auditory stimulation sequence; The steady-state analysis module is configured to extract a first data segment corresponding to the steady-state evoked segment from the continuous scalp EEG signal, extract steady-state auditory evoked response features from the first data segment, and calculate a first anesthesia depth index based on the steady-state auditory evoked response features. The cognitive analysis module is configured to extract a second data segment corresponding to the cognitive exploration segment from the continuous scalp EEG signal, process the second data segment using an overlapping signal separation algorithm to obtain an event-related potential waveform, extract time-domain features from the event-related potential waveform, and calculate a second anesthesia depth index based on the time-domain features. The confidence assessment module is configured to determine a first confidence weight for the first anesthesia depth index based on the signal quality of the first data segment, and to determine a second confidence weight for the second anesthesia depth index based on the residual noise level of the second data segment. The fusion output module is configured to use the first confidence weight and the second confidence weight to perform weighted fusion of the first anesthesia depth index and the second anesthesia depth index, and generate and output the anesthesia depth index.

9. An electronic device, characterized in that, The device includes a processor, a memory, a user interface, and a network interface. The memory is used to store instructions. The user interface and the network interface are both used to communicate with other devices. The processor is used to execute the instructions stored in the memory to cause the electronic device to perform the method as described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores instructions that, when executed, perform the method as described in any one of claims 1-7.

Images