Anesthesia depth monitoring method based on brain wave detection

Abstract

The present application relates to the technical field of medical monitoring, in particular to a method for monitoring depth of anesthesia based on electroencephalogram detection, comprising: acquiring multi-lead scalp electroencephalogram, frontal bipolar surface electromyogram data and wake state segment; performing quality screening on the wake state segment to establish a wake baseline data; dividing the electroencephalogram and electromyogram into time slices according to a unified time reference, calculating band power and coherence to obtain electroencephalogram analysis data; determining contact abnormalities and frontal muscle contaminated leads in combination with the multi-source acquisition data, electroencephalogram analysis data and wake baseline data, determining non-contact abnormal leads as candidate leads; determining leads and time slices dominated by cortical electroencephalogram from the candidate leads, and selecting effective cortical leads; calculating electroencephalogram indexes when the number of effective cortical leads meets the requirements, otherwise marking insufficient data state; outputting anesthesia suppression trends and various state prompts according to the electroencephalogram indexes and abnormal states; the present application can display corresponding results according to lead positions and state categories.

Description

Technical Field

[0001] This invention relates to the field of medical monitoring technology, specifically to a method for monitoring the depth of anesthesia based on electroencephalogram (EEG) detection. Background Technology

[0002] Existing methods for monitoring the depth of anesthesia mainly rely on the analysis of patients' electroencephalogram (EEG) signals;

[0003] In related technologies, in order to monitor the patient's depth of anesthesia, a few frontal leads can be used to collect EEG signals and directly calculate the depth of anesthesia index; or, conventional filtering and fixed thresholds can be used to process the EEG signals to remove electromyographic interference and determine the patient's anesthetic status.

[0004] However, in complex clinical settings, EEG signals collected using a few frontal leads are easily interfered with by frontalis muscle activity, resulting in monitoring accuracy lower than the set standard. Conventional filtering and fixed threshold processing of EEG signals make it difficult to identify abnormal electrode contact and easily filter out effective EEG signals. Therefore, the anesthesia depth monitoring methods in related technologies cannot simultaneously improve anti-interference performance and monitoring accuracy. Summary of the Invention

[0005] Existing anesthesia depth monitoring technologies struggle to simultaneously improve anti-interference performance and monitoring accuracy when faced with frontalis muscle activity contamination and abnormal electrode contact identification. Therefore, there is a need for an anesthesia depth monitoring method based on EEG detection that can integrate multi-source EEG and EMG data to eliminate electrode and EMG interference and dynamically extract effective cortical leads. The objective of this invention can be achieved through the following technical solution:

[0006] A method for monitoring the depth of anesthesia based on electroencephalogram (EEG) detection includes: acquiring multi-lead scalp EEG data from the frontal, central, and occipital regions, bipolar surface electromyography data from the frontal region, periodic impedance readings, lead status, and amplifier saturation status as multi-source acquisition data; acquiring segments of the awake state before anesthesia induction; and performing quality screening on the awake state segments to establish awake baseline data.

[0007] The multi-lead scalp EEG and the frontal bipolar surface electromyography are divided into time slices according to a unified time reference. The frequency band power, inter-lead amplitude square coherence, low-frequency power change synchronicity, low-frequency coherence, power peak time difference, EEG first-order differential amplitude, and the proportion of sampling points in amplifier saturation state are calculated to obtain EEG analysis data.

[0008] Based on the multi-source acquisition data, the EEG analysis data, and the awake baseline data, contact abnormality leads and time slices are determined, and non-contact abnormality leads are identified as candidate leads; based on the multi-source acquisition data, the EEG analysis data, and the awake baseline data, frontalis muscle contamination leads and time slices and the degree of frontalis muscle contamination are determined; based on the candidate leads, cortical EEG-dominant leads and time slices are determined, and effective cortical leads are selected;

[0009] When the number of effective cortical leads is not less than the preset number, EEG indicators are calculated within a rolling time window; otherwise, a data insufficiency status is marked. Based on the EEG indicators, abnormal contact leads and time slices, frontalis muscle contamination leads and time slices, and the degree of frontalis muscle contamination, the following outputs are provided: anesthesia inhibition trend, EEG recovery trend, frontalis muscle contamination indication, abnormal contact leads and the location of the abnormal contact leads on the scalp, the number of effective cortical leads, and a data insufficiency status indication.

[0010] Preferably, the acquisition of the multi-source acquisition data includes: acquiring the multi-lead scalp electroencephalogram (EEG) with a signal frequency band of 0.5–80 Hz and a sampling rate of 250–500 Hz, and the frontal bipolar surface electromyography (EMG) with a bandwidth of 20–150 Hz;

[0011] The multi-lead scalp EEG and the frontal bipolar surface electromyography are aligned with a unified timestamp so that the multi-lead scalp EEG and the frontal bipolar surface electromyography use the same time slice number; the periodic impedance readings are obtained according to a preset detection cycle, and the periodic impedance readings of the previous detection cycle are used between two adjacent impedance readings;

[0012] Preferably, the calculation of the EEG analysis data includes: using 1 second as a time slice and a 2-second analysis window centered on the current time slice; dividing the frequency bands of the multi-lead scalp EEG and the frontal bipolar surface electromyography into 0.5–4 Hz, 4–8 Hz, 8–13 Hz, 13–20 Hz and 20–80 Hz;

[0013] Calculate the synchronicity of low-frequency power changes in the range of 0.5–30 Hz, the coherence of low-frequency power changes in the range of 0.5–30 Hz, the peak power time difference in the range of 0.5–4 Hz, the first-order differential amplitude of the EEG, and the proportion of sampling points in amplifier saturation.

[0014] The synchronicity of the low-frequency power change is determined based on the consistency ratio of the 0.5-30Hz low-frequency power change direction of any two leads, and the direction of the low-frequency power change is determined based on the sign of the low-frequency power change amount.

[0015] Preferably, the determination of the frontal muscle contamination leads and time slices includes: calculating the proportion of the 20-80Hz high-frequency power in the frontal leads to the total power of 0.5-80Hz in the multi-lead scalp EEG, and the root mean square value of the frontal bipolar surface electromyography.

[0016] When the proportion is greater than the 75th percentile of the proportion of the high-frequency power of 20-80Hz in the corresponding lead to the total power of 0.5-80Hz in the awake reference data, the root mean square value of the frontal bipolar surface electromyography is greater than the 75th percentile of the root mean square value of the frontal bipolar surface electromyography in the awake reference data, and the amplitude square coherence between the frontal lead and the central or occipital leads in the multi-lead scalp EEG is lower than the median of the amplitude square coherence between the 20-80Hz leads of all candidate lead pairs in the current time slice and the 25th percentile of the amplitude square coherence between the 20-80Hz leads of the corresponding lead pair in the awake reference data, the corresponding frontal lead and the time slice are determined to be in a frontal muscle contamination state.

[0017] Within a preset statistical window, the proportion of the number of time slices in a frontalis muscle contamination state to the total number of time slices in the frontal leads is the degree of frontalis muscle contamination.

[0018] Preferably, the determination of the contact abnormal leads and time slices includes: for each lead and corresponding time slice, calculating the impedance change at adjacent impedance detection times, the proportion of sampling points in amplifier saturation state, the median absolute deviation of the first-order differential amplitude of EEG, and the correlation coefficient between the 0.5-30Hz EEG signal and the EEG signals of spatially adjacent leads within the time slice;

[0019] When the periodic impedance reading is greater than the preset high impedance upper limit threshold, the impedance change is greater than the median impedance in the conscious reference data plus three times the absolute deviation of the median impedance, the lead status shows detachment or short circuit, the proportion of sampling points in amplifier saturation is greater than 5%, or the median absolute deviation is greater than the 99th percentile of the first-order difference amplitude of the EEG in the conscious reference data and the correlation coefficient is less than 0.2, it is determined that the corresponding lead and time slice are in an abnormal contact state.

[0020] Preferably, the determination of the dominant cortical EEG leads and time slices includes: forming candidate lead pairs from two spatially adjacent or pre-paired candidate leads according to anatomical location, and calculating the low-frequency coherence of 0.5–30 Hz, the power peak time difference of 0.5–4 Hz, and the direction of low-frequency power change;

[0021] When the time difference between the peak power of 0.5 to 4 Hz is not greater than 200 ms, the low-frequency coherence is not lower than the median of the low-frequency coherence of all candidate lead pairs in the current time slice, and the low-frequency power change direction of the two candidate leads is consistent, the candidate lead pair is determined to be in a cortical EEG dominant state, and each candidate lead belonging to the candidate lead pair is determined to be a lead in a cortical EEG dominant state.

[0022] Preferably, the selection of effective cortical leads includes: identifying candidate leads that are in a predominant cortical EEG state as effective cortical leads; when the number of effective cortical leads is not less than three, taking the median of the EEG waveform of each effective cortical lead at each sampling point to obtain the effective cortical median waveform; when the number of effective cortical leads is less than three, marking the current time slice as a data insufficiency state.

[0023] Preferably, the establishment of the awake baseline data includes: acquiring a pre-anesthesia induction awake state segment with a duration of 60 to 180 seconds, and dividing the pre-anesthesia induction awake state segment into a 1-second baseline segment;

[0024] The following benchmark segments were excluded: those with more than 5% of sampling points in amplifier saturation, abnormal lead status, frontal lead EEG signal peak-to-peak amplitude exceeding the 99th percentile of the corresponding peak-to-peak amplitude of all benchmark segments, spatially adjacent lead correlation coefficient below 0.3, impedance exceeding the preset impedance range, or EEG first-order difference amplitude exceeding the 99th percentile of the corresponding statistic of all benchmark segments.

[0025] For the remaining reference segments, calculate the root mean square value of EEG, power of each frequency band, impedance, root mean square value of EMG on the frontal bipolar surface, the proportion of power of each frequency band to the total power of 0.5-80Hz, the first-order difference amplitude of EEG, the median and percentile of the square coherence of amplitude between leads of each frequency band, and the median absolute deviation of impedance.

[0026] When the remaining reference segments are not less than 50% of the total number of reference segments, the burst inhibition amplitude threshold is determined based on the 50th percentile of the root mean square value of EEG in the awake reference data; when the remaining reference segments are less than 50% of the total number of reference segments, the corresponding judgment dependent on the awake reference is skipped and the data insufficiency status prompt is output.

[0027] Preferably, the calculation of the EEG indicators includes: the EEG indicators include burst suppression ratio and cortical low-frequency ratio; within a rolling time window of a selected value in length 60-300s, the burst suppression ratio and the cortical low-frequency ratio are calculated based on the effective cortical median waveform, wherein the burst suppression ratio is the ratio of the burst suppression duration to the effective duration after deducting the time corresponding to abnormal contact leads and time slices, amplifier saturation state and data insufficiency state;

[0028] When the root mean square value of the effective cortical median waveform is lower than the burst suppression amplitude threshold When the duration is not less than 0.5s, it is recorded as a low-amplitude state. When there is a root mean square value not lower than the burst suppression amplitude threshold within the range of 4s before the start time to 4s after the end time of the low-amplitude state. When an effective burst segment, at least two of the effective cortical leads are simultaneously in the low-amplitude state and there are no abnormal contact leads, time slices and amplifier saturation states, the actual duration of the low-amplitude state is accumulated as the burst suppression duration.

[0029] The cortical low-frequency ratio is the ratio of the power at 0.5–13 Hz to the power at 13–30 Hz calculated from the median waveform of the effective cortex.

[0030] Preferably, the output includes: updating the most recently calculated burst suppression ratio, cortical low-frequency ratio, frontalis muscle contamination level, number of effective cortical leads, contact abnormal leads and their duration every 10 seconds within the rolling time window, and comparing the value of the current update cycle with the value of the previous update cycle.

[0031] When the and the When the burst suppression ratio and the cortical low-frequency ratio both increase relative to the previous update cycle, the number of effective cortical leads is not less than three, and the root mean square value of the frontal bipolar electromyography (EMG) is not higher than the 75th percentile of the corresponding frontal bipolar EMG root mean square value in the awake baseline data, the anesthesia suppression trend is output, wherein... It is a positive integer greater than 1;

[0032] When the burst suppression ratio decreases, the cortical low-frequency ratio decreases, and the number of effective cortical leads increases or the proportion of effective cortical leads in a cortical EEG-dominant state to the candidate leads increases, the EEG recovery trend is output.

[0033] When the proportion of time slices in the frontal leads within the pre-anesthesia induction awake state segment that are determined to be in a frontalis muscle contamination state is higher than the 75th percentile of the proportion of frontalis muscle contamination state in the awake baseline data, and the root mean square value of the frontal bipolar surface electromyography is higher than the 75th percentile of the corresponding root mean square value of the frontal bipolar surface electromyography in the awake baseline data, a frontalis muscle contamination indication is output; based on the contact abnormal leads and time slices, amplifier saturation status and data insufficiency status, the contact abnormal leads and their scalp locations, the number of effective cortical leads, and the data insufficiency status indication are output.

[0034] Compared with the prior art, the present invention has the following beneficial effects:

[0035] This method acquires multi-source data such as multi-lead scalp EEG and bipolar surface electromyography, and establishes awake baseline data by combining the quality screening of awake state segments. It outputs individualized comparison benchmarks to avoid the deletion of effective signals by fixed thresholds. In the abnormality identification stage, it integrates the calculation of EEG band power and spatial coherence, and superimposes impedance changes and amplifier saturation state for multi-dimensional evaluation to clarify the degree of contact abnormality and frontalis muscle contamination and improve anti-interference ability.

[0036] During the lead selection phase, dynamic selection of cortical dominant leads is achieved based on spatial adjacency and low-frequency power synchronicity, and the burst suppression ratio and cortical low-frequency ratio are output, coupled with individualized benchmark thresholds to improve the accuracy of EEG index calculation.

[0037] During the trend assessment phase, a rolling time window is used with continuous comparisons across update cycles to generate anesthesia suppression trends and EEG recovery trends, ensuring the objectivity and reliability of the monitoring results. Simultaneously, invalid data is effectively removed by relying on insufficient data status markers, and multi-dimensional status warnings are achieved by combining abnormal lead scalp location indicators. Furthermore, continuous monitoring under complex conditions is ensured by retaining effective non-frontal cortical leads, thereby improving the anti-interference capability and monitoring accuracy of anesthesia depth monitoring. Attached Figure Description

[0038] The above and other objects, features and advantages of the present invention will become clearer from the following description of embodiments of the invention with reference to the accompanying drawings, in which:

[0039] Figure 1 A flowchart illustrating the steps of an anesthesia depth monitoring method based on electroencephalogram (EEG) detection, as provided in this application embodiment. Detailed Implementation

[0040] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments.

[0041] Reference Figure 1 A method for monitoring the depth of anesthesia based on electroencephalogram (EEG) detection includes: acquiring multi-lead scalp EEG of the frontal, central and occipital regions, bipolar surface electromyography of the frontal region, periodic impedance readings, lead status and amplifier saturation status as multi-source acquisition data, and acquiring a segment of the awake state before anesthesia induction.

[0042] Quality screening of conscious state segments was conducted to establish conscious baseline data.

[0043] Multi-lead scalp EEG and frontal bipolar surface electromyography are divided into time slices according to a unified time base. The frequency band power, inter-lead amplitude square coherence, low-frequency power change synchronicity, low-frequency coherence, power peak time difference, EEG first-order differential amplitude, and the proportion of sampling points in amplifier saturation state are calculated to obtain EEG analysis data.

[0044] Based on multi-source acquisition data, EEG analysis data, and awake baseline data, contact abnormality leads and time slices were determined, and non-contact abnormality leads were identified as candidate leads; based on multi-source acquisition data, EEG analysis data, and awake baseline data, frontalis muscle contamination leads, time slices, and the degree of frontalis muscle contamination were determined.

[0045] Based on the candidate leads, determine the leads and time slices in which the cortical EEG is dominant, and select the effective cortical leads;

[0046] When the number of effective cortical leads is not less than the preset number, EEG indicators are calculated within the rolling time window; otherwise, the data is marked as insufficient.

[0047] Based on EEG indicators, contact abnormality leads and time slices, frontalis muscle contamination leads and time slices, and the degree of frontalis muscle contamination, the anesthesia inhibition trend, EEG recovery trend, frontalis muscle contamination indications, contact abnormality leads and their scalp locations, the number of effective cortical leads, and data insufficiency indications are output.

[0048] The acquisition of multi-source data includes: multi-lead scalp EEG with a signal frequency band of 0.5–80 Hz and a sampling rate of 250–500 Hz, and frontal bipolar surface electromyography with a bandwidth of 20–150 Hz;

[0049] Align the multi-lead scalp EEG and frontal bipolar surface electromyography with a unified timestamp so that the multi-lead scalp EEG and frontal bipolar surface electromyography use the same time slice number; obtain periodic impedance readings according to a preset detection cycle, and use the periodic impedance readings of the previous detection cycle between two adjacent impedance readings.

[0050] The calculation of EEG analysis data includes: using 1 second as a time slice and a 2-second analysis window centered on the current time slice; dividing the frequency bands of multi-lead scalp EEG and frontal bipolar surface electromyography into 0.5–4 Hz, 4–8 Hz, 8–13 Hz, 13–20 Hz and 20–80 Hz;

[0051] Calculate the synchronicity of low-frequency power changes in the range of 0.5–30 Hz, the coherence of low-frequency power changes in the range of 0.5–30 Hz, the peak power time difference in the range of 0.5–4 Hz, the first-order differential amplitude of the EEG, and the proportion of sampling points in amplifier saturation.

[0052] Among them, the synchronicity of low-frequency power change is determined by the consistency ratio of the direction of low-frequency power change in the 0.5-30Hz range of any two leads, and the direction of low-frequency power change is determined by the sign of the amount of low-frequency power change.

[0053] This embodiment employs an EEG acquisition device, a frontalis muscle surface electromyography acquisition device, an impedance detection circuit, a lead status detection circuit, an amplifier status detection circuit, and a processor; the EEG acquisition device is connected to multiple scalp electrodes located on the patient's forehead, central region, and occipital region, and the frontalis muscle surface electromyography acquisition device is connected to bipolar surface electrodes located on both sides of the frontalis muscle belly.

[0054] The EEG acquisition device outputs EEG sampling values ​​with sampling timestamps, and the frontalis muscle surface electromyography acquisition device outputs frontalis muscle sampling values ​​with the same time reference; the impedance detection circuit outputs the impedance readings of each scalp electrode, the lead status detection circuit outputs the status of detachment, short circuit or input over-range, and the amplifier status detection circuit outputs a mark indicating whether each sampling point is in a saturated state.

[0055] The processor arranges the EEG sampling values ​​and frontalis muscle sampling values ​​according to a unified sampling timestamp; when the sampling times of the two types of sampling data are not completely consistent, the processor classifies them into the same time slice according to the timestamp, so that each EEG time slice corresponds to a frontalis muscle time slice.

[0056] The time slice length is determined to be 1 second according to the embodiment, and the spectrum calculation uses a 2-second analysis window centered on the current time slice; the status update of the current time slice and the spectrum analysis respectively adopt the corresponding data range to keep the time slice number consistent and meet the frequency resolution requirements of the low frequency band;

[0057] The impedance detection circuit reads the impedance according to the detection cycle supported by the device; after receiving a new impedance reading, the processor associates and saves the reading with the corresponding lead and detection cycle number; before the next impedance reading, the processor uses the most recently read impedance reading and its anomaly judgment result for the lead, and does not interpolate the impedance data into a continuous measurement value between two impedance detection times; the time attribute of the impedance data is synchronized with the actual detection cycle.

[0058] After obtaining the awake state segment before anesthesia induction, the processor checks the amplifier saturation ratio, lead status, impedance range, correlation between adjacent leads, and first-order differential amplitude of EEG in time slices.

[0059] The processor retains segments that meet the quality requirements and calculates the root mean square value of EEG, band power, root mean square value of frontalis muscle, band power ratio, amplitude square coherence between leads, first-order difference amplitude of EEG, and impedance statistics for each lead from the retained segments to form awake baseline data.

[0060] If the proportion of effective segments required for subsequent benchmark calculations is insufficient, the processor saves a benchmark shortage flag so that judgments that depend on the awake benchmark are not directly used for numerical output.

[0061] For each EEG lead, the processor applies a Hanning window to the EEG sample values ​​within a 2-second analysis window and performs a Fast Fourier Transform to calculate the discrete power spectral density. The frequency band power is accumulated according to 0.5~4Hz, 4~8Hz, 8~13Hz, 13~20Hz and 20~80Hz respectively;

[0062] For any two leads, the processor calculates the cross power spectrum and amplitude squared coherence based on the spectral values ​​of the two leads. The specific formula for calculating amplitude squared coherence is as follows:

[0063]

[0064] in, For amplitude squared coherence, For frequency, Let x be the cross power spectral density of lead x and y. and , respectively, are the autopower spectral densities of lead x and lead y, where x and y represent any two different leads, and are averaged over the frequency points of the corresponding frequency band;

[0065] For adjacent time slices, the processor compares the low-frequency power changes of two leads from 0.5 to 30 Hz. When the signs of the two changes are the same, it is recorded that the direction of the low-frequency power change of that lead pair is consistent in the current time slice; otherwise, it is recorded that the direction is inconsistent. The processor further calculates the proportion of leads that are consistent in direction to form the low-frequency power change synchronization.

[0066] The processor applies a 0.5–4 Hz bandpass filter to the data in each lead analysis window to extract the low-frequency waveform envelope in the time domain, determines the time when the envelope has the maximum amplitude in the current time slice, and calculates the time difference between the peak envelope values ​​of two leads.

[0067] EEG analysis data is formed by the synchronicity of low-frequency power changes, low-frequency coherence of 0.5–30 Hz, and power peak time difference of 0.5–4 Hz. The processor first marks the contact abnormal leads and time slices based on lead status, impedance, saturation ratio, first-order difference, and correlation between adjacent leads, and then determines the leads that are not marked as contact abnormal as candidate leads.

[0068] For frontal leads, the processor compares the proportion of high-frequency power (20–80 Hz) to the total power (0.5–80 Hz) with the corresponding percentile of the awake baseline, and compares the root mean square (RMS) value of bipolar electromyography (EMG) of the frontalis muscle with its awake baseline. When both the proportion of high-frequency power in the frontal EEG and the RMS value of the frontalis muscle increase simultaneously, and the high-frequency coherence between the frontal leads and the central or occipital leads is lower than the corresponding threshold, the processor marks the corresponding frontal leads and time slices as frontalis muscle contamination. This marking only applies to leads, time slices, and frequency bands that meet the conditions, without changing the original state of other frequency bands.

[0069] In the candidate leads, the processor forms lead pairs according to spatial adjacency or pre-set anatomical location pairings. The pre-set anatomical location pairings are represented as lead spatial adjacency sets, specifically: based on the international 10-20 standard lead system, fixed pairing tuples are formed, including frontal pole 1 and central area 3, frontal pole 2 and central area 4, central area 3 and occipital area 1, and central area 4 and occipital area 2. For each lead pair, the processor compares low-frequency coherence, power peak time difference, and low-frequency power change direction.

[0070] When a lead pair meets the criteria for cortical EEG dominance, the processor marks the candidate leads in the lead pair as cortical EEG dominance leads; the processor counts the number of leads in the current time slice that are marked; when the number is not less than the preset number, the processor continues to calculate the EEG indicators in the rolling time window; otherwise, it outputs a data insufficiency status.

[0071] In this embodiment, the EEG acquisition device, the frontalis muscle surface electromyography acquisition device, and each state detection circuit provide voltage, impedance, and device status data corresponding to the patient's physiological state and electrode contact state.

[0072] After the processor calculates the bandwidth power, coherence, synchronization and quality status of these data, it outputs the anesthesia inhibition trend, EEG recovery trend, frontalis muscle contamination indication, abnormal leads and scalp location, number of effective cortical leads and data insufficiency status to the display device, so that the display device outputs the corresponding results according to the lead location and status category.

[0073] The determination of the frontal muscle contamination leads and time slices includes: calculating the proportion of the high-frequency power of the frontal leads in the 20-80Hz range to the total power in the 0.5-80Hz range in the multi-lead scalp EEG, and the root mean square value of the frontal bipolar surface electromyography.

[0074] When the proportion is greater than the 75th percentile of the proportion of the high-frequency power of 20-80Hz in the corresponding lead to the total power of 0.5-80Hz in the awake baseline data, the root mean square value of the frontal bipolar surface electromyography is greater than the 75th percentile of the root mean square value of the frontal bipolar surface electromyography in the awake baseline data, and the amplitude square coherence between the frontal lead and the central or occipital leads in the multi-lead scalp EEG is lower than the median of the amplitude square coherence between the 20-80Hz leads of all candidate lead pairs in the current time slice and the 25th percentile of the amplitude square coherence between the 20-80Hz leads of the corresponding lead pair in the awake baseline data, the corresponding frontal lead and time slice are identified as frontalis muscle contamination state.

[0075] Within a statistical window with a length selected from 60s to 300s, the proportion of time slices in the frontalis muscle contamination leads to the total number of time slices in the frontalis leads within that statistical window represents the degree of frontalis muscle contamination.

[0076] The determination of the abnormal leads and time slices includes: for each lead and corresponding time slice, calculating the impedance change at adjacent impedance detection times, the proportion of sampling points in amplifier saturation, the median absolute deviation of the first-order differential amplitude of the EEG, and the Pearson correlation coefficient between the 0.5–30 Hz EEG signal within the time slice and the EEG signals of spatially adjacent leads. ;

[0077] When the impedance is greater than If the impedance change is greater than the median impedance plus three times the absolute deviation of the median impedance in the conscious baseline data, the lead status shows detachment or short circuit, the sampling point ratio is greater than 5%, or the median absolute deviation is greater than the 99th percentile of the first-order difference amplitude of the EEG in the conscious baseline data and the correlation coefficient is less than 0.2, the corresponding lead and time slice are determined to be in an abnormal contact state.

[0078] The determination of the dominant cortical EEG leads and time slices includes: forming candidate lead pairs from two spatially adjacent or pre-paired candidate leads according to anatomical location, and calculating the low-frequency coherence of 0.5–30 Hz, the power peak time difference of 0.5–4 Hz, and the direction of low-frequency power change;

[0079] When the time difference between the peak power of 0.5 to 4 Hz is no greater than 200 ms, the low-frequency coherence is no less than the median of the low-frequency coherence of all candidate lead pairs in the current time slice, and the direction of low-frequency power change of the two candidate leads is consistent, the candidate lead pair is determined to be in the cortical EEG dominant state, and each candidate lead belonging to the candidate lead pair is determined to be a lead in the cortical EEG dominant state.

[0080] The selection of effective cortical leads includes: identifying candidate leads that are in a predominantly cortical EEG state as effective cortical leads;

[0081] When there are at least three effective cortical leads, the median of the EEG waveforms of each effective cortical lead is taken at each sampling point to obtain the median waveform of the effective cortex; when there are fewer than three effective cortical leads, the current time slice is marked as insufficient data.

[0082] In this embodiment, the processor calculates the high-frequency power of 20-80Hz and the total power of 0.5-80Hz from the analysis window data of each frontal lead, and uses the ratio of the two as the proportion of frontal high-frequency power; the processor calculates the root mean square value of the electromyography sampling value output by the frontal bipolar surface electromyography acquisition device; the processor compares the two results with the 75th percentile of the corresponding awake benchmark, and only when both results exceed their respective benchmark values ​​will the spatial coherence judgment continue to be performed;

[0083] The processor extracts 20-80 Hz coherence from all candidate lead pairs in the current time slice and calculates the median of these coherences. For each frontal lead, the processor also extracts 20-80 Hz coherence between the frontal lead and the central or occipital leads and compares it with the median of coherence of candidate lead pairs in the current time slice and the 25th percentile of coherence of the corresponding lead pair in the wakefulness baseline.

[0084] When the coherence of the frontal lead is lower than both of the above comparison values ​​at the same time, the processor records the frontal lead and the current time slice as a frontalis muscle contamination state.

[0085] The degree of frontalis muscle contamination is determined by the processor counting the number of time slices marked as frontalis muscle contamination in the frontal leads within a preset statistical window, and then dividing by the total number of time slices in the frontal leads. If the root mean square value of frontal bipolar surface electromyography (EMG) is elevated but the proportion of high-frequency power in frontal EEG does not exceed the 75th percentile of the awake baseline, the processor will not identify the time slice as frontalis muscle contamination. Instead, it will retain the changes in frontalis muscle reference signal and the corresponding EEG data to avoid making a judgment of frontalis muscle contamination based solely on the elevation of EMG reference signal.

[0086] For contact anomaly detection, the processor calculates the first-order difference of EEG in the current time slice of each lead; the difference between the EEG values ​​of adjacent sampling points is formed into a first-order difference sequence, and the median absolute deviation of the sequence is used to represent the degree of waveform change; the processor calculates the impedance change based on the most recent impedance reading and the previous impedance reading corresponding to the current time slice; between adjacent impedance readings, the current time slice uses the impedance value and status saved in the previous detection cycle.

[0087] When the processor detects that the impedance exceeds the preset high impedance limit, or the impedance change exceeds the median of the wake reference impedance plus three times the absolute deviation of the median impedance, it directly marks the corresponding lead as a contact abnormality candidate; if the lead status detection circuit outputs a disconnection or short circuit state, the processor also marks the lead as a contact abnormality.

[0088] For amplifier saturation, the processor calculates the ratio of the number of saturated sampling points to the total number of sampling points in the current time slice; when this ratio is greater than 5%, it is marked as a candidate for contact anomaly.

[0089] When the median absolute deviation of the first-order difference amplitude exceeds the 99th percentile of the corresponding statistic of the awake benchmark, the processor continues to calculate the correlation coefficient of 0.5 to 30 Hz between the lead and spatially adjacent leads; only when the correlation coefficient is lower than 0.2 is the lead and time slice identified as contact abnormality; EEG activity with a first-order difference amplitude exceeding the corresponding preset threshold and a correlation coefficient of not less than 0.2 with adjacent leads is preserved;

[0090] The processor removes leads identified as having contact abnormalities from the candidate lead set; for candidate leads not identified as having contact abnormalities, the processor forms lead pairs according to spatial adjacency or pre-paired anatomical location; the processor calculates the coherence of each lead pair from 0.5 to 30 Hz, determines the direction of the low-frequency power change of the two leads from 0.5 to 30 Hz, and determines the time difference of the power peaks of the two leads from 0.5 to 4 Hz.

[0091] The processor first calculates the median of low-frequency coherence for all candidate lead pairs in the current time slice, and then compares the low-frequency coherence of each lead pair with this median. When the peak power time difference of the lead pair is not greater than 200ms, the low-frequency coherence is not lower than the median of the coherence of the candidate lead pairs in the current time slice, and the low-frequency power change direction of the two leads is consistent, the processor determines the lead pair as the cortical EEG dominant state, and marks the two candidate leads in the lead pair as cortical EEG dominant leads.

[0092] The processor counts the number of leads marked as dominated by cortical EEG; when there are no fewer than three valid cortical leads, the processor sorts the EEG values ​​of the same sampling point in each valid cortical lead according to the sampling point number and takes the median value to form the median waveform of the valid cortical EEG.

[0093] When the number of effective cortical leads is less than three, the processor does not generate the effective cortical median waveform for burst suppression calculation, but instead marks the current time slice as a data insufficiency state; this state is saved with the time slice number and used as a validity condition for the numerical results in subsequent trend outputs;

[0094] The establishment of the awake baseline data includes: acquiring a 60-180s segment of the awake state before anesthesia induction, and dividing the 60-180s segment of the awake state before anesthesia induction into a 1s baseline segment;

[0095] Sampling points excluded if the proportion of sampling points in amplifier saturation exceeds 5%, or if the lead status is abnormal, or if the peak-to-peak amplitude of the EEG signal in the frontal lead exceeds the 99th percentile of the corresponding peak-to-peak amplitude of all reference segments, or if the correlation coefficient between spatially adjacent leads is less than 0.3, or if the impedance is not specified. ~ A reference segment whose first-order difference amplitude in EEG exceeds the 99th percentile of the corresponding statistic for all reference segments within the range;

[0096] For the remaining reference segments, calculate the root mean square value of EEG, power of each frequency band, impedance, root mean square value of EMG on the frontal bipolar surface, the proportion of power of each frequency band to the total power of 0.5-80Hz, the first-order difference amplitude of EEG, the median and percentile of the square coherence of amplitude between leads of each frequency band, and the median absolute deviation of impedance.

[0097] When the remaining baseline segments are no less than 50% of the total number of baseline segments, the 50th percentile of the root mean square value of EEG in the conscious baseline data is recorded as _____. And determine the burst suppression amplitude threshold. When the remaining reference fragments are less than 50% of the total number of reference fragments, skip the corresponding judgment based on the awake reference and output a data shortage status prompt.

[0098] The calculation of EEG indicators includes: EEG indicators include burst suppression ratio and cortical low frequency ratio; within a rolling time window of selected values ​​in length 60 to 300 seconds, the burst suppression ratio and cortical low frequency ratio are calculated based on the effective cortical median waveform. The burst suppression ratio is the ratio of burst suppression duration to the effective duration after deducting the time corresponding to abnormal contact leads and time slices, amplifier saturation state and data insufficiency state.

[0099] When the root mean square value of the effective cortical median waveform is lower than the burst suppression amplitude threshold If the duration is not less than 0.5s, it is recorded as a low-amplitude state. When there is a root mean square value not lower than the burst suppression amplitude threshold within the range of 4s before the start time to 4s after the end time of the low-amplitude state. When the effective burst segment, at least two effective cortical leads are simultaneously in a low-amplitude state and there are no abnormal contact leads, time slices and amplifier saturation states, the actual duration of the low-amplitude state is accumulated as the burst suppression duration;

[0100] The cortical low-frequency ratio is the ratio of the power in the 0.5–13 Hz range to the power in the 13–30 Hz range calculated from the median waveform of the effective cortex.

[0101] The output includes: updated every 10 seconds of the most recently calculated burst suppression ratio, cortical low-frequency ratio, frontalis muscle contamination level, number of effective cortical leads, contact aberration leads, and time slices within the rolling time window, and the [number of]th [time slices] is also included. The value of the update cycle is related to the first... The values ​​of the update cycle are compared, among which... It is a positive integer greater than 1;

[0102] When the and the When the burst suppression ratio and cortical low-frequency ratio of the update cycle are both increased compared to the previous update cycle, the number of effective cortical leads is not less than three, and the root mean square value of frontal bipolar surface electromyography is not higher than the 75th percentile of the corresponding root mean square value of frontal bipolar surface electromyography in the awake baseline data, the output anesthesia suppression trend is indicated.

[0103] When the burst suppression ratio decreases, the cortical low-frequency ratio decreases, and the number of effective cortical leads increases or the proportion of effective cortical leads in the candidate leads that are in a cortical EEG-dominant state increases, the output EEG shows a recovery trend.

[0104] When the proportion of time slices in the frontal leads within the pre-anesthesia induction period that are identified as having frontal muscle contamination is higher than the 75th percentile of the proportion of frontal muscle contamination in the awake baseline data, and the root mean square value of frontal bipolar surface electromyography is higher than the 75th percentile of the corresponding root mean square value of frontal bipolar surface electromyography in the awake baseline data, a frontal muscle contamination warning is output.

[0105] Based on the abnormal contact leads and time slice, amplifier saturation status, and insufficient data status, the output includes the abnormal contact leads, the scalp location of the abnormal contact leads, the number of effective cortical leads, and a prompt indicating insufficient data.

[0106] In this embodiment, the processor receives the conscious EEG, frontalis bipolar surface electromyography, impedance readings, lead status, and amplifier saturation status before anesthesia induction. The processor divides the conscious EEG into continuous 1-second reference segments according to the reference segment number, and calculates the root mean square value, peak-to-peak value, first-order difference amplitude of the EEG, power of each frequency band, amplitude square coherence between leads, impedance, and frontalis muscle root mean square value for each lead within each reference segment.

[0107] The processor first calculates temporary statistics for anomaly screening of all benchmark segments, including peak-to-peak values, EEG first-order difference amplitude, and distribution statistics of correlation coefficients between adjacent leads; then, based on the temporary statistics, lead status, amplifier saturation status, and impedance range, abnormal benchmark segments are eliminated.

[0108] After the elimination is completed, the processor only uses the retained baseline fragments to establish formal awake baseline data, so that abnormal fragments will not continue to participate in the statistics of the patient's own threshold; for the screening condition that requires the determination of the 99th percentile according to all baseline fragments, all baseline fragments refer to the baseline fragments that have completed time slice division and have corresponding statistics. Temporary statistics are calculated once within the same awake state fragment and remain unchanged.

[0109] The processor checks the amplifier saturation point ratio of each reference segment; when the saturation point ratio exceeds 5%, the processor deletes the reference segment; if the lead status shows detachment, short circuit or other abnormalities, the processor also deletes the reference segment; for the frontal lead, the processor compares the peak-to-peak value of the current segment with the 99th percentile of the corresponding peak-to-peak value of all reference segments.

[0110] If the value exceeds the percentile, the segment is deleted; the processor also calculates the correlation coefficient between spatially adjacent leads, and deletes the corresponding baseline segment when the correlation coefficient is below 0.3; when the impedance exceeds the preset impedance range, or the first-order difference amplitude of the EEG exceeds the 99th percentile of the corresponding statistic of all baseline segments, the corresponding baseline segment is also deleted.

[0111] The processor performs statistical analysis on the retained baseline segments to obtain the 50th percentile of the root mean square value of EEG for each lead, the median and percentile of the band power, the 75th percentile of the proportion of high-frequency power in the frontal lobe, the 50th and 75th percentiles of the root mean square value of the frontalis muscle, the median impedance, the absolute deviation of the median impedance, the 99th percentile of the first-order difference amplitude of EEG, and the corresponding percentile of inter-lead coherence.

[0112] The above statistical values ​​were saved as baseline data for awakeness and used as comparative values ​​for patients in subsequent assessments of contact abnormalities, frontalis muscle contamination, and burst inhibition.

[0113] When the remaining baseline segments are no less than 50% of the total number of baseline segments, the processor extracts the 50th percentile of the root mean square value of the awake baseline EEG, denoted as . and according to Determine the threshold for burst suppression amplitude;

[0114] The processor compares the fixed minimum amplitude and the scaling value of the wakefulness reference amplitude, and uses the larger of the two values ​​as the threshold. When the remaining reference segments are less than 50% of the total number of reference segments, the processor does not calculate the burst suppression threshold and corresponding numerical indicators that depend on the reference, but outputs a data shortage status prompt to the display device.

[0115] During real-time monitoring, the processor generates effective cortical median waveforms from no fewer than three effective cortical leads; for each analysis segment, the processor calculates the root mean square value of the effective cortical median waveform.

[0116] When the root mean square value is lower than When the continuous duration is not less than 0.5s, the processor forms a low-amplitude state candidate segment; the processor continues to check the range from 4s before the start time to 4s after the end time of this candidate segment, and searches for a root mean square value not lower than within this range. Effective burst fragments;

[0117] A valid burst segment is defined as one in which there are no contact anomalies, amplifier saturation, or insufficient data markers at the corresponding sampling point, and at least two of the valid cortical leads have a root mean square value not lower than [value missing]. The duration of an effective burst segment must cover at least one adjacent analysis segment at the start or end boundary of the candidate low-amplitude state to avoid identifying a single anomalous sampling point as a burst segment.

[0118] The processor also checks whether at least two valid cortical leads are synchronously in a low-amplitude state during the low-amplitude state, and checks whether there are contact abnormalities or amplifier saturation in the corresponding leads; only when the low-amplitude duration meets the conditions, there are valid burst segments in the preceding and following ranges, at least two valid cortical leads are synchronously in a low-amplitude state and there are no contact abnormalities or amplifier saturation, does the processor determine the low-amplitude state as a burst suppression segment, and accumulates its actual duration as the burst suppression duration.

[0119] For times marked as contact abnormality, amplifier saturation, or insufficient data, the processor does not include them in the burst suppression duration or the effective duration; between adjacent burst suppression segments, if no waveform meeting the conditions for an effective burst segment appears within the interval, the processor treats them as a continuous part of the same candidate low-amplitude state; if an effective burst segment meeting the conditions appears within the interval, it is accumulated according to the actual time boundaries of the low-amplitude state and the effective burst segment respectively.

[0120] The processor calculates the burst suppression duration and effective duration within a selected rolling time window; the effective duration is equal to the total duration of the rolling time window minus the duration corresponding to contact anomalies, amplifier saturation, and insufficient data; the burst suppression ratio is equal to the ratio of burst suppression duration to effective duration;

[0121] When the effective duration is less than half of the total duration of the rolling time window, or when the number of effective cortical leads is less than three, the processor does not output the numerical burst suppression ratio of the current window, but outputs a data shortage status; when there is no effective duration within the rolling time window, the processor does not perform the burst suppression ratio division calculation and marks the time window as a data shortage status.

[0122] For the effective cortical median waveform, the processor calculates the power in the 0.5–13 Hz range and the power in the 13–30 Hz range, and divides the former by the latter to obtain the cortical low-frequency ratio. When the power in the 13–30 Hz range is zero or lower than the effective power range required for calculation, the processor processes the data according to the numerical stability condition of the effective power and retains the data validity mark to avoid outputting an unfounded ratio when there is insufficient data.

[0123] Specifically, when the power in the 13-30Hz range does not meet the effective power condition, the processor does not output the cortical low-frequency ratio of the current time window and marks the index corresponding to that time window as insufficient data.

[0124] The processor updates the burst suppression ratio, cortical low-frequency ratio, frontalis muscle contamination level, number of effective cortical leads, contact abnormal leads and their duration within the rolling time window according to a 10-second update cycle; the processor saves the results of the current update cycle and the previous update cycle, and compares the direction of change of consecutive update cycles;

[0125] When the burst suppression ratio and cortical low-frequency ratio both increase relative to their respective previous update cycles in two consecutive update cycles, and the number of effective cortical leads is not less than three and the root mean square value of bipolar surface electromyography of the frontalis muscle is not higher than the 75th percentile of the awake baseline, the processor outputs the anesthesia suppression trend to the display device.

[0126] In terms of the timing of trend determination, the processor considers each 10-second update cycle as a cycle in which data processing has been completed. When the nth update cycle is completed, the processor only saves the burst suppression ratio, cortical low-frequency ratio and related states of that cycle, and compares them with the results of the (n-1)th update cycle. After the n+1th update cycle is completed, the results of the n+1th update cycle are compared with those of the nth update cycle, and the direction of continuous change is determined by combining the results of the previous comparison.

[0127] Where n is the sequence number of the current completed update cycle, n-1 is the sequence number of the previous update cycle, and n+1 is the sequence number of the next update cycle. The sequence numbers increase sequentially according to the 10-second update cycle.

[0128] Therefore, in the embodiment, the comparison between the nth and n+1th update cycles is completed after the data of the n+1th update cycle is actually obtained, and the output is the trend confirmation based on the obtained data, without predicting the update cycle that has not yet arrived.

[0129] When any indicator in the nth, n+1th, or their respective previous update period is marked as insufficient data, the processor does not output the corresponding trend based on the direction of change of that indicator until it obtains data for consecutive update periods that meet the validity conditions.

[0130] When the burst suppression ratio decreases, the cortical low-frequency ratio decreases, and the number of effective cortical leads increases, or the proportion of effective cortical leads in a cortical EEG-dominant state to candidate leads increases, the processor outputs an EEG recovery trend. This trend judgment uses effective cortical leads that have passed the screening for contact abnormalities and frontal muscle contamination. Therefore, even when frontal leads are marked as contamination or abnormal, trend results can still be formed from effective cortical leads in the central or occipital regions.

[0131] For frontalis muscle contamination alerts, the processor calculates the proportion of time slices in the waking state segment where the frontal leads are identified as having frontalis muscle contamination and obtains the 75th percentile of this proportion distribution. In real-time monitoring, when the root mean square value of the bipolar surface electromyography of the frontal region is higher than its 75th percentile of the waking baseline, and the comparison result of the proportion of frontalis contamination in the waking state segment meets the conditions of the embodiment, the processor outputs a frontalis muscle contamination alert.

[0132] The prompt and the trend of anesthesia inhibition are output separately; when the frontalis muscle activity increases but the burst inhibition ratio and the cortical low-frequency ratio do not change synchronously, the processor does not output the trend of deepening anesthesia inhibition based solely on the frontalis muscle activity; when the effective awake baseline fragment is insufficient to form the 75th percentile of the proportion of frontal contamination, the processor does not perform the percentile comparison and outputs a data insufficiency status prompt.

[0133] For contact abnormalities, amplifier saturation, and insufficient data, the processor sends the abnormal lead number and its pre-associated scalp location to the display device, along with the number of valid cortical leads and the data insufficiency status; the display device displays the corresponding lead location, duration, and numerical index according to the received status category.

[0134] When the frontal leads cannot be used for effective cortical calculation, but the central or occipital region still has no less than three effective cortical leads, the processor continues to use the central or occipital region leads to form the effective cortical median waveform and retains the frontal contamination or contact abnormality information as an independent output.

[0135] In this embodiment, the length of the rolling time window is selected from 60 to 300 seconds during device initialization; when no other configuration instructions are received, 120 seconds is used as the default length and remains unchanged during the same monitoring process; when the selected window length has not been accumulated at the beginning of the monitoring phase, the data that is insufficient for the complete window is not extrapolated, but the window is marked as having insufficient data.

[0136] The burst suppression ratio is output only after the window has accumulated a complete and effective duration of at least half of the total window duration; each update uses the most recent complete rolling time window obtained up to the 10-second update cycle, and the window advances forward in time, without changing its length due to a single abnormal time slice;

[0137] Among them, the effective power condition of 13-30Hz power is determined according to the quantization resolution of the equipment acquisition channel and the awake reference noise; the processor counts the 5th percentile of the 13-30Hz power of each lead in the awake reference segment, and takes the larger value between this value and the power noise floor measured by the amplifier in the input short-circuit state as the lower limit of effective power.

[0138] When the power of the 13-30Hz range in the real-time window is greater than the effective power lower limit, it is determined that the effective power condition is met; otherwise, it is marked as insufficient data. When the device does not provide an input short-circuit noise floor, the 5th percentile of the power of the 13-30Hz range in the wakeful reference segment is directly used as the effective power lower limit. When the lower limit is zero, it is still not used as a basis for calculation, but the corresponding window is marked as insufficient data.

[0139] For the 75th percentile of the proportion of frontal contamination in the conscious state segment, the processor takes the proportion of time slices in each 1-second baseline segment in which all frontal leads are marked as frontalis muscle contamination as a statistical sample, and takes the sample value at the 75th percentile after arranging the samples from smallest to largest.

[0140] When the sample size is less than 4, the percentile is not established and the data shortage status is output; for other percentiles based on the sobriety benchmark, the same sample sorting rules are used; when the effective sample size is insufficient to form the corresponding statistic, the judgment on which the statistic depends is skipped.

[0141] When the burst suppression ratio, cortical low-frequency ratio, or the number of effective cortical leads used for trend judgment is marked as insufficient data in any update cycle, the processor only saves the state of that cycle and does not use it as an increasing or decreasing value in trend comparison; only when the relevant indicators in the two adjacent comparison cycles have effective values ​​and the number of effective cortical leads is not less than three, the anesthesia suppression trend or EEG recovery trend is determined according to the embodiment.

[0142] The above specific embodiments are merely specific implementations of the present invention. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for monitoring the depth of anesthesia based on electroencephalogram (EEG) detection, characterized in that, include: Acquire multi-lead scalp EEG, bipolar surface electromyography of the frontal region, central region, and occipital region, periodic impedance readings, lead status, and amplifier saturation status as multi-source acquisition data, and acquire segments of the awake state before anesthesia induction. The conscious state segments were subjected to quality screening to establish conscious baseline data; The multi-lead scalp EEG and the frontal bipolar surface electromyography are divided into time slices according to a unified time reference. The frequency band power, inter-lead amplitude square coherence, low-frequency power change synchronicity, low-frequency coherence, power peak time difference, EEG first-order differential amplitude, and the proportion of sampling points in amplifier saturation state are calculated to obtain EEG analysis data. Based on the multi-source acquisition data, the EEG analysis data, and the awake baseline data, contact abnormal leads and time slices are determined, and non-contact abnormal leads are identified as candidate leads; based on the multi-source acquisition data, the EEG analysis data, and the awake baseline data, frontalis muscle contamination leads and time slices and the degree of frontalis muscle contamination are determined. Based on the candidate leads, determine the leads and time slices where cortical EEG is dominant, and select the effective cortical leads; When the number of effective cortical leads is not less than the preset number, EEG indicators are calculated within the rolling time window; otherwise, the data is marked as insufficient. Based on the aforementioned EEG indicators, contact abnormality leads and time slices, frontalis muscle contamination leads and time slices, and the degree of frontalis muscle contamination, the output shows the anesthesia inhibition trend, EEG recovery trend, frontalis muscle contamination indications, contact abnormality leads and scalp location of contact abnormality leads, number of effective cortical leads, and data insufficiency indications.

2. The method for monitoring the depth of anesthesia based on electroencephalogram (EEG) detection according to claim 1, characterized in that, The acquisition of the multi-source acquisition data includes: acquiring the multi-lead scalp electroencephalogram (EEG) with a signal frequency band of 0.5–80 Hz and a sampling rate of 250–500 Hz, and the frontal bipolar surface electromyography (EMG) with a bandwidth of 20–150 Hz. The multi-lead scalp EEG and the frontal bipolar surface electromyography are aligned with a unified timestamp so that the multi-lead scalp EEG and the frontal bipolar surface electromyography use the same time slice number; the periodic impedance readings are obtained according to a preset detection cycle, and the periodic impedance readings of the previous detection cycle are used between two adjacent impedance readings.

3. The method for monitoring anesthesia depth based on electroencephalogram (EEG) detection according to claim 1, characterized in that, The calculation of the EEG analysis data includes: using 1 second as a time slice and a 2-second analysis window centered on the current time slice; dividing the frequency bands of the multi-lead scalp EEG and the frontal bipolar surface electromyography into 0.5–4 Hz, 4–8 Hz, 8–13 Hz, 13–20 Hz and 20–80 Hz; Calculate the synchronicity of low-frequency power changes in the 0.5–30 Hz range, the coherence of low-frequency power changes in the 0.5–30 Hz range, the time difference of power peak values ​​in the 0.5–4 Hz range, the first-order differential amplitude of the EEG, and the proportion of sampling points in amplifier saturation. The synchronicity of low-frequency power changes is determined based on the consistency ratio of the 0.5–30 Hz low-frequency power change directions of any two leads, and the direction of low-frequency power change is determined based on the sign of the low-frequency power change.

4. The method for monitoring anesthesia depth based on electroencephalogram (EEG) detection according to claim 1, characterized in that, The determination of the frontal muscle contamination leads and time slices includes: calculating the proportion of the high-frequency power of the frontal leads in the multi-lead scalp EEG to the total power in the range of 0.5-80Hz, and the root mean square value of the frontal bipolar surface electromyography. When the proportion is greater than the 75th percentile of the proportion of the high-frequency power of 20-80Hz in the corresponding lead to the total power of 0.5-80Hz in the awake reference data, the root mean square value of the frontal bipolar surface electromyography is greater than the 75th percentile of the root mean square value of the frontal bipolar surface electromyography in the awake reference data, and the amplitude square coherence between the frontal lead and the central or occipital leads in the multi-lead scalp EEG is lower than the median of the amplitude square coherence between the 20-80Hz leads of all candidate lead pairs in the current time slice and the 25th percentile of the amplitude square coherence between the 20-80Hz leads of the corresponding lead pair in the awake reference data, the corresponding frontal lead and the time slice are determined to be in a frontal muscle contamination state. Within a preset statistical window, the proportion of the number of time slices in a frontalis muscle contamination state to the total number of time slices in the frontal leads represents the degree of frontalis muscle contamination.

5. The method for monitoring anesthesia depth based on electroencephalogram (EEG) detection according to claim 1, characterized in that, The determination of the contact abnormal leads and time slices includes: for each lead and corresponding time slice, calculating the impedance change at adjacent impedance detection times, the proportion of sampling points in amplifier saturation state, the median absolute deviation of the first-order differential amplitude of EEG, and the correlation coefficient between the 0.5-30Hz EEG signal and the EEG signal of spatially adjacent leads within the time slice. When the periodic impedance reading is greater than the preset high impedance upper limit threshold, the impedance change is greater than the median impedance in the conscious reference data plus three times the absolute deviation of the median impedance, the lead status shows detachment or short circuit, the proportion of sampling points in amplifier saturation is greater than 5%, or the median absolute deviation is greater than the 99th percentile of the first-order differential amplitude of the EEG in the conscious reference data and the correlation coefficient is less than 0.2, the corresponding lead and time slice are determined to be in an abnormal contact state.

6. The method for monitoring anesthesia depth based on electroencephalogram (EEG) detection according to claim 1, characterized in that, The determination of the dominant cortical EEG leads and time slices includes: forming candidate lead pairs from two spatially adjacent or pre-paired candidate leads according to anatomical location, and calculating the low-frequency coherence of 0.5–30 Hz, the power peak time difference of 0.5–4 Hz, and the direction of low-frequency power change; When the time difference between the peak power of 0.5 to 4 Hz is not greater than 200 ms, the low-frequency coherence is not lower than the median of the low-frequency coherence of all candidate lead pairs in the current time slice, and the low-frequency power change direction of the two candidate leads is consistent, the candidate lead pair is determined to be in a cortical EEG dominant state, and each candidate lead belonging to the candidate lead pair is determined to be a lead in a cortical EEG dominant state.

7. The method for monitoring anesthesia depth based on electroencephalogram (EEG) detection according to claim 6, characterized in that, The selection of effective cortical leads includes: identifying the candidate leads that are in a predominantly cortical EEG state as effective cortical leads; When there are at least three effective cortical leads, the median of the EEG waveforms of each effective cortical lead is taken at each sampling point to obtain the median effective cortical waveform; when there are fewer than three effective cortical leads, the current time slice is marked as insufficient data.

8. The method for monitoring the depth of anesthesia based on electroencephalogram (EEG) detection according to claim 7, characterized in that, The establishment of the awake baseline data includes: acquiring a pre-anesthesia induction awake state segment with a duration of 60 to 180 seconds, and dividing the pre-anesthesia induction awake state segment into a 1-second baseline segment; The following benchmark segments were excluded: those with more than 5% of sampling points in amplifier saturation, abnormal lead status, frontal lead EEG signal peak-to-peak amplitude exceeding the 99th percentile of the corresponding peak-to-peak amplitude of all benchmark segments, spatially adjacent lead correlation coefficient below 0.3, impedance exceeding the preset impedance range, or EEG first-order difference amplitude exceeding the 99th percentile of the corresponding statistic of all benchmark segments. For the remaining reference segments, calculate the root mean square value of EEG, power of each frequency band, impedance, root mean square value of EMG on the frontal bipolar surface, the proportion of power of each frequency band to the total power of 0.5-80Hz, the first-order difference amplitude of EEG, the median and percentile of the square coherence of amplitude between leads of each frequency band, and the median absolute deviation of impedance. When the remaining baseline segments are no less than 50% of the total number of baseline segments, the 50th percentile of the root mean square value of the EEG in the awake baseline data is recorded as _____. And determine the burst suppression amplitude threshold. When the remaining reference fragments are less than 50% of the total number of reference fragments, skip the corresponding determination based on the awake reference and output the data shortage status prompt.

9. The method for monitoring the depth of anesthesia based on electroencephalogram (EEG) detection according to claim 8, characterized in that, The calculation of the EEG indicators includes: the EEG indicators include burst suppression ratio and cortical low frequency ratio; Within a rolling time window with a length of 60 to 300 seconds, the burst suppression ratio and the cortical low-frequency ratio are calculated based on the effective cortical median waveform. The burst suppression ratio is the ratio of the burst suppression duration to the effective duration after deducting the time corresponding to abnormal contact leads and time slices, amplifier saturation state and data insufficiency state. When the root mean square value of the effective cortical median waveform is lower than the burst suppression amplitude threshold When the duration is not less than 0.5s, it is recorded as a low-amplitude state. When there is a root mean square value not lower than the burst suppression amplitude threshold within the range of 4s before the start time to 4s after the end time of the low-amplitude state. When an effective burst segment, at least two of the effective cortical leads are simultaneously in the low-amplitude state and there are no abnormal contact leads, time slices and amplifier saturation states, the actual duration of the low-amplitude state is accumulated as the burst suppression duration. The cortical low-frequency ratio is the ratio of the power at 0.5–13 Hz to the power at 13–30 Hz calculated from the median waveform of the effective cortex.

10. The method for monitoring the depth of anesthesia based on electroencephalogram (EEG) detection according to claim 9, characterized in that, The output includes: updating every 10 seconds the most recently calculated burst suppression ratio, cortical low-frequency ratio, frontalis muscle contamination level, number of effective cortical leads, contact aberration leads, and time slice within the rolling time window, and setting the [number]th [time slice] as [value]. The value of the update cycle is related to the first... The values ​​of the update cycle are compared, among which... It is a positive integer greater than 1; When the and the When the burst suppression ratio and the cortical low-frequency ratio both increase in the update cycle relative to their respective previous update cycle, the number of effective cortical leads is not less than three, and the root mean square value of the frontal bipolar surface electromyography is not higher than the 75th percentile of the corresponding root mean square value of the frontal bipolar surface electromyography in the awake baseline data, the anesthesia suppression trend is output. When the burst suppression ratio decreases, the cortical low-frequency ratio decreases, and the number of effective cortical leads increases or the proportion of effective cortical leads in a cortical EEG-dominant state to the candidate leads increases, the EEG recovery trend is output. When the proportion of time slices in the frontal leads within the pre-anesthesia induction awake state segment that are identified as frontalis muscle contamination is higher than the 75th percentile of the proportion of frontalis muscle contamination in the awake baseline data, and the root mean square value of the frontal bipolar surface electromyography is higher than the 75th percentile of the corresponding root mean square value of the frontal bipolar surface electromyography in the awake baseline data, a frontalis muscle contamination prompt is output. Based on the abnormal contact leads and time slice, amplifier saturation status, and insufficient data status, the output includes the abnormal contact leads and their scalp locations, the number of effective cortical leads, and a prompt indicating insufficient data.

Images