Method and equipment for extracting oxygenation state of brain through noninvasive brain wave in surgical anesthesia

Abstract

The invention discloses a brain wave cleaning method and equipment for extracting a brain oxygenation state in operation anesthesia maintenance through noninvasive brain waves. The system collects five-electrode non-invasive brain wave data under the action of anesthetic for a period of time, and decomposes the change components of brain waves under the action of sedation and analgesia of drugs in brain waves of left and right brains by using algorithms such as wavelet analysis, spectrum analysis and mode recognition, and the decomposition algorithm is based on the coordination characteristic of the left and right brains to realize the analysis of the brain waves of the left and right brains. And dynamic orthogonal bias of left and right brain electroencephalogram components under space and time resolutions is adopted to obtain the change of association strength of left and right brain neural information pathway interaction. Meanwhile, through a decomposition algorithm, brain wave characteristics caused by electric or magnetic interference in surgical anesthesia and high and low frequency differences before and after electroencephalogram recovery of left and right brains after brain wave electromagnetic interference are captured, and influence factors of comprehensive change of the neuron working state caused by unilateral hypoxia are obtained. According to the monitoring equipment, brain wave effective components are recombined on a multi-level surface, a real-time and effective data sequence containing oxygen synthesis components is formed, the time sequence is combined, brain wave basic data containing the oxygenation state after the influence of anesthetic drugs and operation of an operation electrotome are removed are obtained, and the brain oxygenation state is formed.

Description

Technical Field

[0001] The present invention belongs to the technical field of medical devices. Background Art

[0002] In the field of clinical medicine, surgical treatment is an important means of disease intervention. The anesthesia process of surgery is the basis for ensuring the smooth execution of the operation. Surgical anesthesia requires accurate measurement of the functional status and vital sign status of each system of the body. At present, the measurement of vital sign status covers the functions of most important organs and changes in biochemical indicators. Only the measurement requirement for cerebral perfusion is still in its infancy. Perioperative stroke is a serious complication with high disability and fatality rates, occupying a large amount of medical and rehabilitation resources. The incidence of cerebral ischemia in patients after cardiac surgery is 1.6% - 9.7%, and the death risk increases by 7 - 21 times. In 2020, the total number of hospitalizations for cardiovascular and cerebrovascular diseases in China was 24.28 million person-times. Measurement techniques for cerebral hypoxia include: transcranial Doppler ultrasound (TCD), which requires professional personnel for interpretation, and about 10% - 20% of patients do not have a suitable temporal window for monitoring. Somatosensory evoked potential (SEP) has poor specificity, is invasive, and has high professional requirements. Non-invasive cerebral oxygen saturation (NIRS) mainly monitors the frontal lobe rather than the whole brain, and its results do not have absolute value repetition, are easily affected by ambient light, skin, and muscle blood vessels, and the results also vary greatly. Jugular venous blood-brain oxygen saturation is invasive, cannot be continuously monitored, and has high operation requirements. At present, the monitoring of cerebral hypoxia during the perioperative period has a high equipment configuration threshold, strong operation professionalism, lagging indicators, and many interference factors. It is difficult to detect cerebral hypoxia in a timely manner under general anesthesia, which is directly related to serious anesthesia complications such as ischemic stroke, postoperative delirium, and postoperative cognitive impairment. A monitoring device with the function of real-time measurement of cerebral hypoxia and prognosis judgment during the perioperative period has significant significance.

[0003] Starting from the result measurement of the change in the functional state of brain neurons caused by cerebral hypoxia, using the electroencephalogram technology of neuroelectrophysiology, the electroencephalogram contains characteristic components of the functional changes caused by cerebral hypoxia, which is one of the feasible new modalities for real-time measurement of cerebral hypoxia. Electroencephalogram measurement has advantages such as high compliance, strong continuity, clear results, strong computability, high sensitivity, etc., and is the key technology for realizing real-time evaluation and prediction of postoperative complications related to brain diseases during clinical surgical anesthesia. It is also one of the supporting forces leading to the progress of brain science technology. Summary of the Invention

[0004] The purpose of the present invention is to provide a brain-computer interface device that extracts the cerebral oxygenation state through non-invasive electroencephalogram to solve the real-time measurement of the cerebral oxygenation state during the surgical anesthesia process and the major application of predicting the occurrence of postoperative brain diseases including stroke and brain cognition.

[0005] After cerebral hypoxia, it is inevitable to cause changes in the neurological function state of the brain. Electroencephalogram (EEG) is the external information directly reflecting the neurological function state of the brain. Cerebral hypoxia is the cause, and the functional change is the result. The two have an inheritance relationship. By using neuroelectrophysiological techniques and reading the basic characteristics of EEG, the formation of static and dynamic changes can reflect the changes in the neurological function state of the brain, and thus directly reflect the changes in cerebral hypoxia.

[0006] A brain-computer interface device for extracting the cerebral oxygenation state through non-invasive EEG includes: a five-electrode non-invasive EEG acquisition front-end device; a five-electrode wired non-invasive EEG acquisition sensor; a system computing main computer; a display unit and an input keyboard (soft and hard). By wearing an EEG acquisition sensor on the head, the EEG signals on the surface of the frontal skin are collected and transmitted to the EEG acquisition front-end device. The signals are filtered, amplified, and analog-to-digital converted, and then transmitted to the computer in a wired manner. After the computer receives the discrete EEG signals, the wavelet analysis algorithm is used to perform interference decomposition and two-compartment decomposition algorithms on the EEG to obtain interference signal components including eye movement and skin electromyogram tremors, as well as cortical EEG and subcortical EEG components. For these independent signals, multiple combination algorithms are used again, including wavelet, spectral analysis, pattern recognition, etc. algorithms, to obtain the time-shift data sequence parameters of each waveform at each scale in each space (cortex, subcortex) of the left and right brains, representing the intensity values of information interaction between the left and right brains. At the time resolution, based on the effects of anesthetic, sedative, analgesic drugs, and muscle relaxant drugs, the neural channels for external and internal information input are inhibited. In this case, if the cerebral oxygen supply is in a normal state, the dynamic changes of the time-shift data sequences of the two parts of the left and right brains are coordinated and synchronized. This coordination and synchronization is defined as the orthogonal information interaction state of information interaction between the left and right brains. By calculating the dynamic deviation degree of the time-shift data parameter sequences at each scale in each space of the left and right brains, the dynamic orthogonal lateralization value of the EEG components at the spatial and time resolutions of the left and right brains is obtained, reflecting the change in the coordination and synchronization of neural information interaction between the left and right brains after cerebral hypoxia. Thus, a quantitative index related to the cerebral oxygenation state: the cerebral hypoxia index is extracted.

[0007] The calculation of dynamic time-shift data mainly uses the wavelet formula:

[0008]

[0009] (a,τ) is the scaling factor and translation parameter of wavelet transform;

[0010] ω is the angular frequency of wavelet transform;

[0011] ψ(a ω ) is the mother function of wavelet transform;

[0012] X(ω) is the wavelet transform result sequence of the original discrete EEG sequence;

[0013] WT is the time-domain signal sequence after inverse transformation;

[0014] For a certain set of collected EEG data vectors, real-time calculation and processing are performed. Using the multi-scale filter bank algorithm, the wavelet basis functions at each scale window are decomposed. For a set of wavelet basis functions and data at each scale, after the discrete wavelet inverse transformation, a set of time-domain reconstruction functions is obtained. Each reconstruction function represents the performance of brain waves, eye movement waves, and muscle waves at different scales; the scale also corresponds to the frequency components of the signal, distributed in the conventional rhythm and high-frequency rhythm of the electroencephalogram. Select the brain wave reconstruction function corresponding to a certain scale as the spatial distribution calculation sequence, and a set of reconstructed spatial distribution brain wave sequences is obtained. For the left and right brains, they can be marked as two sets of brain wave sequences. The two sequence sets jointly construct a three-dimensional spatial architecture with layered left and right brains at the spatial resolution. The relevant calculation formulas are as follows:

[0015] Discrete EEG data sequence:

[0016] b i1 (t) = [y1 y2 y3…y m-2 y m-1 y m )

[0017] i1: The number of EEG leads;

[0018] m: The number of EEG data;

[0019] y: EEG data sequence;

[0020] t: Time point;

[0021] Wavelet basis function

[0022] (Wb(2^i, w i (r))) i∈z

[0023] (Wb(2^0, w o (r))), (Wb(2^1, w1(r)))…(Wb(2^N, w N (r)))

[0024] w i (r): Represents a set of result data sequences after wavelet transformation;

[0025] r: Represents the serial number in the data sequence;

[0026] z: Time-domain space;

[0027] i: Scale;

[0028] Time-domain reconstruction function:

[0029] f i(t) = ∑Wb(2^i, w i ) * Ψ 2^i (u)

[0030] i: scale;

[0031] Wb: wavelet basis function;

[0032] w i : result data after wavelet transform at each scale;

[0033] Ψ 2^0 (u), Ψ 2^1 (u)... Ψ 2^N (u): mother function scale wavelet data points;

[0034] N: order;

[0035] u: data sequence number of scale wavelet data points;

[0036] t: time point;

[0037] Spatial distribution calculation sequence:

[0038] Fs i (t) ∈ {f i (t) >= A}

[0039] i: scale;

[0040] A: scale threshold, a constant;

[0041] t: time series;

[0042] f i (t): reconstruction function set;

[0043] Fs i (t): electroencephalogram of spatial distribution;

[0044] For the left and right brains, it can be marked as two groups of electroencephalogram sequences:

[0045] Fs_r i (t) ∈ {f_r i (t) < A}

[0046] Fs_l i (t) ∈ {f_l i (t) < A}

[0047] i: scale;

[0048] t: time series;

[0049] r: right brain lead;

[0050] l: left brain lead;

[0051] A: Scale threshold;

[0052] For the two sets of electroencephalogram data sequences of the left and right brains at the above scales, remove the data sequences at the highest and lowest scales, and continue to use the wavelet analysis algorithm for the electroencephalogram data at the remaining scales to obtain two sets of reconstructed electroencephalogram sequence data for the left and right brains at multiple scales after subdivision, covering the gradient distribution of the cortex and subcortex. For each sequence data of the reconstructed electroencephalogram function decomposed above, use the variation value calculation algorithm in the waveform recognition algorithm to extract the variation points of each data sequence:

[0053] Y ik (h) ∈ {Fs_r i (t), Fs_l i (t)};

[0054] Y: Variation point vector matrix;

[0055] i: Scale;

[0056] k: Representing the left and right brains

[0057] h: Time serial number of characteristic data;

[0058] t: Time series;

[0059] Vector matrix Y ik The characteristic data in (h) contains variation points calculated from the basic algorithm:

[0060] Data sequence:

[0061] y(x, i) = {(ff i (j))′, (ff i (j)) * Δt}

[0062] j: Subscript of time discrete data;

[0063] ′: Derivative function

[0064] i: Scale;

[0065] x: Sequence point;

[0066] Δt: Time increment;

[0067] ff: Function in the F series reconstruction function set

[0068] For the variation value matrix Y ik (h), calculate two independent matrices for the left and right brains, respectively forming two-dimensional spatial submatrices, and the differential variable of the variation point is calculated as follows:

[0069] U ik (j, t) = {Δh, Y ik (h) - Yik (h + 1)}

[0070] i: Scale;

[0071] j: Serial number;

[0072] k: 0 - 1, left and right brains;

[0073] h: Variable;

[0074] Δh: Time difference;

[0075] t: Time;

[0076] Dataset U ik (j, t) forms the dynamic layer dataset of the left and right brains, reflecting the time - domain dynamic changes of the reconstructed left and right brainwaves after wavelet transform. Regarding the influence of anesthetic drugs, for the balanced distribution of the left and right brains, the change in oxygenation, associated with local hypoxia, is the reason for the change in the working state of neurons in the left and right brains, which in turn affects the neuron firing process. Calculate the difference at the time resolution of the left - right brain difference matrix to obtain the time - shift calculation formula for the left and right brains:

[0077] Z i (t) = {U i0 (j, t) - U i1 (j, t), Δh}

[0078] i: Scale;

[0079] t: Time point;

[0080] Δh: Time difference;

[0081] j: Serial number;

[0082] Time - shift function Z i (t) represents the subtle changes in the time - shift of the EEG signals at each scale gradient of the left and right brains. Select the integral of the time - shift function within 5 seconds:

[0083] G(t) = ∫Z i (t)^2 * Δt

[0084] i: Scale

[0085] Δt: Time increment

[0086] t: Time series

[0087] Obtain the power magnitude G of the time - shift function, defined as the dynamic ortholateral value, reflecting the change in the coordination synchronization of the neural information interaction between the left and right brains after cerebral hypoxia.

[0088] For the two sets of EEG data sequences of the left and right brain at each scale, the data sequences of the highest and lowest scales excluded above are regarded as the main components of myoelectric, electrical and oculomotor interference. The power spectrum algorithm is used to reconstruct the function for this part:

[0089]

[0090] ff i (t): highest scale and lowest scale wavelet reconstruction function;

[0091] t: time point;

[0092] ω1: angular frequency of the power spectrum;

[0093] i: 0-1, highest scale and lowest scale;

[0094] j1: data window value for power spectrum calculation;

[0095] dt: time increment

[0096] The power spectrum data containing electromagnetic, static, electric knife and other interference components and subtle EEG components can be obtained, including:

[0097] P ik ={α1, β1, δ1, θ1, sef1, mef1};

[0098] i: 0-1, highest scale and lowest scale;

[0099] k: 0-1, left and right brain

[0100] α1, β1, θ1, θ1: power percentage of each band;

[0101] sef1: edge frequency for power spectrum calculation;

[0102] mef1: center frequency of power spectrum calculation;

[0103] The four sets of data mark the power spectrum characteristics of the lowest and highest frequencies. The power ratio of the two frequency bands at the two scales is calculated over time. The calculation formula is as follows:

[0104] B ik (t) = {(α1 + β1) ik / (δ1,θ1) ik ∈t};

[0105] i: 0-1, highest scale and lowest scale;

[0106] k: 0-1, left and right brain

[0107] t: time series

[0108] As a signal fluctuation gradient function that changes over time. Combine the numerical values of the edge frequency and the center frequency that change over time:

[0109] SEF1 ik (t)

[0110] MEF1 ik (t)

[0111] i: 0 - 1, the highest scale and the lowest scale;

[0112] k:,, 0 - 1, left and right brains

[0113] t: time series

[0114] Calculate the difference sum of each parameter of the gradient function of the left and right brains to form a gradient difference curve, calculate the area under the curve, and obtain:

[0115] D(t) = {ΔB ik (t), ΔSEF1 ik (t), ΔMEF1 ik (t)}

[0116] A(t) = ∫D(t)^2 * Δt

[0117] Δ: the difference of each parameter between the left and right brains

[0118] Δt: time increment

[0119] The numerical value A represents the difference between the left and right brains of the effective high-frequency and low-frequency signals retained after removing most of the interferences, and obtains the influence component of the comprehensive change of the neuron working state caused by local hypoxia. The numerical values G and A are normalized:

[0120] Cerebral hypoxia index(t) = (exp(G + A) / C) × 100

[0121] C: matching constant

[0122] t: time series

[0123] Obtain a quantitative real-time cerebral hypoxia index of 0 - 100.

[0124] Advantages of the present invention:

[0125] The present invention utilizes neuroelectrophysiological techniques, uses a five-electrode electroencephalogram (EEG) acquisition sensor, and through real-time acquisition of EEG signals, discrete signal processing techniques of wavelet calculation, and a two-compartment EEG analysis method. By analyzing the changes in brain function states, the hypoxic characteristic components related to the nerve function state in the EEG are extracted, directly reflecting the degree of brain hypoxia. It is a new modality for quantitative measurement of hypoxia. Among them, the multi-scale dynamic time-shift algorithm of EEG is a process of brain-like information processing, providing a monitoring device for real-time brain oxygenation status during surgical anesthesia for whole-brain hypoxia monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0126] Figure 1 This is the EEG signal acquisition circuit of the present invention.

[0127] Figure 2 This is the communication connection relationship between the various parts of the 5-electrode EEG information acquisition and calculation unit of the present invention.

[0128] Figure 3 The calculation control circuit of the calculation unit.

[0129] Figure 4 The power management circuit of the device DETAILED DESCRIPTION OF THE INVENTION

[0130] For the principle and structure of the present invention, refer to Figure 1 、 2 as shown. A brain-computer interface device that extracts the ischemic and hypoxic states of the brain through non-invasive EEG. During surgical anesthesia, the changes in brain nerve function caused by brain hypoxia in patients are manifested in the extraction and expression of characteristic components in the EEG. Using the EEG as the original signal to process data, it is discretely collected into a computer, and various mathematical calculation algorithms are applied. The EEG is decomposed into spatial components of the cortex, subcortex, left brain, and right brain, as well as time components that change over time. The time-shifted signals are calculated and extracted for each scale component of the EEG. Through the differential changes in the time shift of the left and right brains, the hypoxia index of the brain is obtained, which is used to measure the degree of hypoxia formed by the changes in the brain oxygenation status due to various reasons such as the influence of anesthetic drugs, surgical stimulation, and fluid intake and output during surgical anesthesia.

[0131] Using an electroencephalogram (EEG) sensor, which is worn on the forehead and ear parts of the head, to non-invasively collect the potential signals of multiple parts of the brain. The signals enter the computing unit via a preamplifier, a multi-stage integration circuit, and an analog-to-digital converter. The signal bandwidth includes the ultra-high frequency components in the brain waves (brain wave components above 30 Hz). The EEG acquisition leads are set to include at least five electrodes in the prefrontal lobes of the left and right brains. The EEG signals go through pre-amplification and analog-to-digital conversion, enter the single-chip computer system, and then, after data encryption and compression, are directly transmitted to the computer unit via a wired connection. The dedicated software system of the computer performs real-time processing, calculation, display, and storage on the received EEG signals. Among them, the calculation part uses algorithms such as wavelet analysis, spectral analysis, and pattern recognition to decompose the various effective components and artifact components related to scale, frequency, and time domain in the brain waves, and combines them into a new data stream. Then, through various algorithms such as time shift calculation and analysis of the data streams of the left and right brains, the characteristic data is extracted.

[0132] The present invention relates to the manufacture of a five-electrode brain wave sensor and a device, an information acquisition and transmission terminal, and related fixed terminals, calculation and data management software, etc., including the acquisition and processing part of multi-channel brain waves, and the extraction and expression of the characteristic components of brain waves for real-time measurement of the brain oxygenation state by a computer.

[0133] Through the five electrodes (bioelectric signal sensors) of the brain wave sensor on the head of the patient population, the multi-channel brain wave signals of the user are collected and transmitted to the filtering, noise control, and amplification input part of the preamplifier circuit, and are respectively sent into the analog-to-digital conversion circuits of the single-chip computer control part of the terminal through the corresponding channels. The single-chip computer control part obtains the digital brain wave (including the high-frequency brain waves above 30 Hz) data. After the data processing process of encryption and compression, the processed data stream is obtained and sent into the dynamic data link buffer queue. The dynamic data link buffer queue is a variable data storage and output structure. According to the computer communication and speed requirements and buffer requirements, the structure of the data in the storage area is different. After the single-chip microcomputer control circuit is triggered by the interruption event of the communication state change, it controls the different permutations and combinations of the collected data. Under the control of the write instruction, the data is written into the storage queue. The data in the queue is transmitted to the communication port of the computer through the communication module of the data port under the control of the read instruction of the single-chip control circuit. The output function of the EEG signal is completed.

[0134] The data transmission, status reception, and command reception of the EEG acquisition terminal are the fundamental guarantees for the practical application of the system and one of the key links for the safe use of the system. When using a nearby computer as the processing platform, the terminal data is directly received through the wifi or Bluetooth communication ports configured on the computer, and all the work of data calculation, display, and storage is completed by the computer's operating system and the brain hypoxia measurement application software. Utilizing the flow control ability of the single-chip computer, the status of data transmission and the size of the internal buffer are processed in real time. The real-time process of network communication is automatically recognized and self-corrected, and a dynamic data link queue area of up to six minutes is set up to ensure the integrity of data and improve the fault tolerance ability. The data transmission of the system adopts an encryption algorithm for data transmission, including the decomposition calculation of dynamic data combinations.

[0135] For the processing of brain waves, discretization processing is performed with a sampling frequency of 1400 / s, a sampling time window of 1.25 s, and a sampling accuracy of 10-bit bite. Algorithms such as wavelet analysis, spectral analysis, and pattern recognition are used to calculate the reconstructed sequences at different scales of the brain and the time-domain characteristics and multi-scale complexity characteristics of each sequence, obtaining the calculation result sequence of the primary processing calculation. Among the brain waves collected bilaterally from the frontal region, there are electrooculogram components and frontalis muscle electrical components. In the wavelet processing calculation, these two bioelectrical signal sequences can be decomposed and used as marker characteristics of signal intensity and interference signals. The wavelet formula is used:

[0136]

[0137] (a,τ), which are the scaling factor and translation parameter of the wavelet transform;

[0138] ω, which is the angular frequency of the wavelet transform;

[0139] ψ(a ω ), which is the mother function of the wavelet transform;

[0140] X(ω) is the wavelet transform result sequence of the original brain wave discrete sequence;

[0141] WT is the time-domain signal sequence after the inverse transform;

[0142] For a certain set of collected EEG data vectors, real-time calculation and processing are performed. Using the multi-scale filter bank algorithm, the wavelet basis functions under each scale window are decomposed. For a set of wavelet basis functions and data at each scale, after the inverse discrete wavelet transform, a set of time-domain reconstruction functions is obtained. Each reconstruction function represents the performance of brain waves, eye movement waves, and muscle waves at different scales; the scale also corresponds to the frequency components of the signal, distributed in the conventional rhythm and high-frequency rhythm of the electroencephalogram. Select the brain wave reconstruction function corresponding to a certain scale as the spatial distribution calculation sequence to obtain a set of reconstructed spatial distribution brain wave sequences. For the left and right brains, they can be marked as two sets of brain wave sequences. The two sets of sequences together construct a three-dimensional spatial architecture with layered left and right brains at the spatial resolution. The relevant calculation formulas are as follows:

[0143] Discrete EEG data sequence:

[0144] b i1 (t) = [y1 y2 y3…y m-2 y m-1 y m

[0145] i1: The number of EEG leads;

[0146] m: The number of EEG data;

[0147] y: EEG data sequence;

[0148] t: Time point;

[0149] Wavelet basis function

[0150] (Wb(2^i, w i (r)))i∈z

[0151] (Wb(2^0, w0(r))), (Wb(2^1, w1(r)))…(Wb(2^N, w N (r)))

[0152] w i (r): Represents a set of result data sequences after wavelet transform;

[0153] r: Represents the serial number in the data sequence;

[0154] z: Time domain space;

[0155] i: Scale;

[0156] Time-domain reconstruction function:

[0157] f i (t) = ∑Wb(2^i, w i )*Ψ 2^i (u) ​

[0158] i: scale;

[0159] Wb: wavelet basis function;

[0160] w i : result data after wavelet transform at each scale;

[0161] Ψ 2^0 (u), Ψ 2^1 (u)... Ψ 2^N (u): mother function scale wavelet data points;

[0162] N: order;

[0163] u: data sequence number of scale wavelet data points;

[0164] t: time point;

[0165] Spatial distribution calculation sequence:

[0166] Fs i (t) ∈ {f i (t) >= A}

[0167] i: scale;

[0168] A: scale threshold, which is a constant;

[0169] t: time series;

[0170] f i (t): reconstruction function set;

[0171] Fs i (t): electroencephalogram of spatial distribution;

[0172] For the left and right brains, they can be marked as two groups of electroencephalogram sequences:

[0173] Fs_r i (t) ∈ {f_r i (t) < A}

[0174] Fs_li(t) ∈ {f_l i (t) < A}

[0175] i: scale;

[0176] t: time series;

[0177] r: right brain lead;

[0178] l: left brain lead;

[0179] A: scale threshold;

[0180] For the two sets of electroencephalogram data sequences of the left and right brains at the above-mentioned scales, remove the data sequences at the highest and lowest scales, and continue to use the wavelet analysis algorithm for the electroencephalogram data at the remaining scales to obtain two sets of reconstructed electroencephalogram sequence data for the left and right brains at multiple scales after subdivision, covering the gradient distribution of the cortex and subcortex. For each sequence data of the electroencephalogram reconstruction function decomposed above, use the variance value calculation algorithm in the waveform recognition algorithm to extract the variance points of each data sequence:

[0181] Y ik (h) ∈ {Fs_r;(t), Fs_l i (t)};

[0182] Y: Variance point vector matrix;

[0183] i: Scale;

[0184] k: Represents the left and right brains

[0185] h: Time serial number of characteristic data;

[0186] t: Time series;

[0187] Vector matrix Y ik (h) The characteristic data in contains the variance point calculation from the basic algorithm:

[0188] Data sequence:

[0189] y(x, i) = {(ff i (j))′, (ff i (j)) * Δt}

[0190] j: Subscript of time discrete data;

[0191] ′: Derivative function

[0192] i: Scale;

[0193] x: Sequence point;

[0194] Δt: Time increment;

[0195] ff: Function in the F series reconstruction function set

[0196] For the variance value matrix Y ik (h), calculate two independent matrices for the left and right brains, respectively form two-dimensional spatial submatrices, and the differential variable of the variance point is calculated as follows:

[0197] U ik (j, t) = {Δh, Y ik (h) - Y ik (h + 1)}

[0198] i: Scale;

[0199] j: Serial number;

[0200] k: 0 - 1, left and right brains;

[0201] h: Variable;

[0202] Δh: Time difference;

[0203] t: Time;

[0204] Dataset U ik (j, t) forms the dynamic layer dataset of the left and right brains, reflecting the time - domain dynamic changes of the reconstructed left and right brainwaves after wavelet transform. Regarding the influence of anesthetic drugs, for the balanced distribution of the left and right brains, the change in oxygenation, associated with local hypoxia, is the reason for the change in the working state of neurons in the left and right brains, and further affects the neuron discharge process. Calculate the difference at the time resolution of the left - right brain difference matrix to obtain the time - shift calculation formula for the left and right brains:

[0205] Z i (t) = {U i0 (j, t) - U i1 (j, t), Δh}

[0206] i: Scale;

[0207] t: Time point;

[0208] Δh: Time difference;

[0209] j: Serial number;

[0210] Time - shift function Z i (t) represents the subtle changes in the time - shift of the electroencephalogram signals at each scale gradient of the left and right brains. Select the integral of the time - shift function within 5 seconds:

[0211] G(t) = ∫Z i (t)^2 * Δt

[0212] i: Scale

[0213] Δt: Time increment

[0214] t: Time series

[0215] Obtain the power magnitude G of the time - shift function, defined as the dynamic orthogonality lateral value, reflecting the change in the coordination synchronization of the information interaction between the left and right brain nerves after brain hypoxia.

[0216] For the two - group electroencephalogram data sequences of the left and right brains at each scale, the data sequences of the highest and lowest scales excluded previously are regarded as the main components of electromyogram, electrical, and electro - oculogram interferences. For the reconstruction function of this part, adopt the power spectrum algorithm:

[0217]

[0218] ff i (t): highest scale and lowest scale wavelet reconstruction function;

[0219] t: time point;

[0220] ω1: angular frequency of the power spectrum;

[0221] i: 0-1, highest scale and lowest scale;

[0222] j1: data window value for power spectrum calculation;

[0223] dt: time increment

[0224] The power spectrum data containing electromagnetic, static, electric knife and other interference components and subtle EEG components can be obtained, including:

[0225] P ik ={α1, β1, δ1, θ1, sef1, mef1};

[0226] i: 0-1, highest scale and lowest scale;

[0227] k: 0-1, left and right brain

[0228] α1, β1, δ1, θ1: power percentage of each band;

[0229] sef1: edge frequency for power spectrum calculation;

[0230] mef1: center frequency of power spectrum calculation;

[0231] The four sets of data mark the power spectrum characteristics of the lowest and highest frequencies. The power ratio of the two frequency bands at the two scales is calculated over time. The calculation formula is as follows:

[0232] B ik (t) = {(α1 + β1) ik / (δ1, O1) ik ∈t};

[0233] i: 0-1, highest scale and lowest scale;

[0234] k: 0-1, left and right brain

[0235] t: time series

[0236] As a function of the gradient of the signal fluctuation over time. Combined with the values of the edge frequency and the center frequency over time:

[0237] SEF1 ik (t)

[0238] MEF1 ik (t)

[0239] i: 0 - 1, the highest and lowest scales;

[0240] k:,, 0 - 1, left and right brains

[0241] t: time series

[0242] Calculate the difference sum of each parameter of the left and right brain gradient functions to form a gradient difference curve, and calculate the area under the curve to obtain:

[0243] D(t) = {ΔB ik (t), ΔSEF1 ik (t), ΔMEF1 ik (t)}

[0244] A(t) = ∫D(t)^2 * Δt

[0245] Δ: the difference of each parameter between the left and right brains

[0246] Δt: time increment

[0247] The value A represents the left and right brain differences of the effective high - frequency and low - frequency signals retained after removing most interferences, and obtains the influencing component of the comprehensive change of the neuron working state caused by local hypoxia. The values G and A are normalized by:

[0248] Cerebral hypoxia index(t) = (exp(G + A) / C) × 100

[0249] C: matching constant

[0250] t: time series

[0251] Obtain a quantitative real - time cerebral hypoxia index of 0 - 100.

Claims

1. A brain-computer interface system for extracting the cerebral oxygenation state during surgical anesthesia through non-invasive brain waves, characterized in that, The system includes: a two-channel non-invasive electroencephalogram acquisition front-end device; a five-electrode wired non-invasive electroencephalogram acquisition sensor; a system computing main computer; a display unit and an input keyboard (soft and hard); by wearing a disposable electroencephalogram acquisition sensor on the head, the electroencephalogram signals on the surface of the frontal skin are collected and transmitted to the electroencephalogram acquisition front-end device. The signals are filtered, amplified, and analog-to-digital converted, and then transmitted to the computer in a wired manner. After the computer receives the discrete electroencephalogram signals, the wavelet analysis algorithm is used to perform interference decomposition and two-compartment decomposition algorithms on the electroencephalogram to obtain interference signal components such as eye movement and skin myoelectric tremors, as well as cortical electroencephalogram and subcortical electroencephalogram components. For these independent signals, various combination algorithms are used again, including wavelet, spectral analysis, pattern recognition, etc. algorithms, to obtain the time-shift data sequence parameters of each waveform at each scale in each space (cortex, subcortex) of the left and right brains, representing the intensity value of information interaction between the left and right brains. At the time resolution, based on the effects of anesthesia, sedation, analgesia drugs, and muscle relaxant drugs, the neural channels for external and internal information input are inhibited. In this case, if the cerebral oxygen supply is in a normal state, the dynamic changes of the time-shift data sequences of the two parts of the left and right brains are coordinated and synchronous. This coordinated synchronization is defined as the orthogonal information interaction state of information interaction between the left and right brains. By calculating the dynamic deviation degree of the time-shift data parameter sequences at each scale in each space of the left and right brains, the dynamic orthogonal lateralization value of the electroencephalogram components at the spatial and temporal resolutions of the left and right brains is obtained, reflecting the change in the coordinated synchronization of neural information interaction between the left and right brains caused by cerebral hypoxia. Thus, a quantitative index related to the oxygenation state of the brain: the cerebral hypoxia index is extracted. The calculation of dynamic time-shift data mainly uses the wavelet formula: (a, τ), which are the scaling factor and translation parameter of the wavelet transform; ω, which is the angular frequency of the wavelet transform; ψ(a ω ), is the mother function of wavelet transform; X(ω) is the wavelet transform result sequence of the original discrete electroencephalogram sequence; WT is the time-domain signal sequence after the inverse transform; For a certain group of electroencephalogram data vector groups collected, real-time calculation and processing are performed, and the wavelet basis functions at each scale window are decomposed by the multi-scale filter bank algorithm. For a group of wavelet basis functions and data at each scale, after the discrete wavelet inverse transform, a group of time-domain reconstruction functions are obtained. Each reconstruction function represents the performance of electroencephalogram, eye movement electroencephalogram, and myoelectric wave at different scales; the scale also corresponds to the frequency components of the signal, distributed in the conventional rhythm and high-frequency rhythm of the electroencephalogram; select the electroencephalogram reconstruction function corresponding to a certain scale as the spatial distribution calculation sequence to obtain a group of reconstructed spatial distribution electroencephalogram sequences. For the left and right brains, they can be marked as two groups of electroencephalogram sequences. The two sequence sets jointly construct a three-dimensional spatial architecture with layered left and right brains at the spatial resolution. The relevant calculation formulas are as follows: Discrete electroencephalogram data sequence: b i1 (t) = [y1 y2 y3 … y m-2 y m-1 y m ​ i1: the number of electroencephalogram leads; m: the number of electroencephalogram data; y: the electroencephalogram data sequence; t: the time point; Wavelet basis function (Wb(2^i, w i (r))) i∈z (Wb(2^0, w0(r))), (Wb(2^1, w1(r)))…(Wb(2^N, w N (r))) w i (r): represents a set of result data sequences after wavelet transform; r: representing the serial number in the data sequence; z: the time-domain space; i: the scale; Time-domain reconstruction function: f i (t) = ∑Wb(2^i, w i ) * Ψ 2^i (u) i: the scale; Wb: the wavelet basis function; w i : Result data after wavelet transform at each scale; Ψ 2^0 (u), Ψ 2^1 (u)...Ψ 2^N (u): Mother function scale wavelet data point; N: the order; u: the data serial number of the scale wavelet data point; t: Time point; Spatial distribution calculation sequence: Fs i (t) ∈ {f i (t) >= A} i: Scale; A: Scale threshold, which is a constant; t: Time series; f i (t): set of reconstruction functions; Fs i (t): spatially distributed brain waves; For the left and right brains, it can be marked as two groups of electroencephalogram sequences: Fs_r i (t) belongs to {f_r i (t) < A} Fs_l i (t) belongs to {f_l i (t) < A} i: Scale; t: Time series; r: Right brain lead; l: Left brain lead; A: Scale threshold; For the two groups of electroencephalogram data sequences of the left and right brains at the above-mentioned scales, remove the data sequences of the highest and lowest scales, and continue to use the wavelet analysis algorithm for the electroencephalogram data of the remaining scales to obtain two sets of reconstructed electroencephalogram sequence data of the left and right brains at multiple scales after subdivision, covering the gradient distribution of the cortex and subcortex. For each sequence data of the reconstructed function of the electroencephalogram decomposed above, use the variance value calculation algorithm in the waveform recognition algorithm to extract the variance points of each data sequence: Y ik (h) ∈ {Fs_r i (t), Fs_l i (t)}; Y: Variance point vector matrix; i: Scale; k: Representing the left and right brains h: Time serial number of feature data; t: Time series; Vector matrix Y ik The characteristic data in (h) contains mutation points calculated from the basic algorithm: Data sequence: y(x, i) = {(ff i (j))′, (ff i (j)) * Δt} j: Subscript of time discrete data; ′: Derivative function i: Scale; x: Sequence point; Δt: Time increment; ff: Function in the F series reconstruction function set For the variant value matrix Y ik (h), calculate two independent matrices for the left and right brains, respectively forming two-dimensional spatial sub-matrices. The differential variable of the variant point is calculated as follows: U ik (j, t) = {Δh, Y ik (h) - Y ik (h + 1)} i: Scale; j: Serial number; k: 0 - 1, left and right brains; h: Variable; Δh: Time difference; t: Time; Dataset U ik (j, t) forms the dynamic layer dataset of the left and right brains, reflecting the time-domain dynamic changes of the reconstructed left and right brainwaves after wavelet transform. Regarding the influence of anesthetic drugs, for the balanced distribution of the left and right brains, the change in oxygenation, associated with local hypoxia, is the reason for the change in the working state of neurons in the left and right brains, which in turn affects the neuronal discharge process. Calculate the difference at the time resolution of the left and right brain difference matrices to obtain the time shift calculation formula for the left and right brains: Z i (t) = {U i0 (j, t) - U i1 (j, t), Δh} i: Scale; t: Time point; Δh: Time difference; j: Serial number; Time-shift function Z i (t) represents the subtle changes in the time shift of EEG signals at various scale gradients of the left and right brains. Select the integral of the time-shift function within 5 seconds: G(t) = ∫Z i (t)^2 * Δt i: Scale Δt: Time increment t: Time series Obtain the power magnitude G of the time-shifted function, defined as the dynamic orthogonal lateral value, which reflects the change in the coordination synchronization of the neural information interaction between the left and right brains after cerebral hypoxia. For the two groups of electroencephalogram data sequences of the left and right brains at each scale, the data sequences of the highest and lowest scales excluded above are regarded as the main components of electromyogram, electrical and electrooculogram interference. For this part of the reconstruction function, use the power spectrum algorithm: ff i (t): highest scale and lowest scale wavelet reconstruction functions; t: Time point; ω1: Angular frequency of the power spectrum; i: 0 - 1, highest and lowest scales; j1: Data window value for power spectrum calculation; dt: Time increment Power spectrum data containing interference components such as electromagnetic, electrostatic, and electrosurgical and fine electroencephalogram components can be obtained, including: P ik = {α1, β1, δ1, θ1, sef1, mef1}; i: 0 - 1, highest and lowest scales; k: 0 - 1, left and right brains α1, β1, δ1, θ1: Power percentages of each band; sef1: Edge frequency calculated by the power spectrum; mef1: Central frequency calculated by the power spectrum; 4 groups of data mark the power spectrum characteristics of the lowest and highest frequencies. Calculate the change of the power ratio of the two frequency bands at two scales over time, and the calculation formula is as follows: B ik (t) = {(α1 + β1) ik / (δ1, θ1) ik ∈ t}; i: 0 - 1, highest and lowest scales; k: 0 - 1, left and right brains t: Time series As a signal fluctuation gradient function that changes over time. Combine the values of the edge frequency and the central frequency that change over time: SEF1 ik (t) MEF1 ik (t) i: 0 - 1, highest and lowest scales; k: 0 - 1, left and right brains t: Time series Calculate the sum of the differences of each parameter of the left and right brain gradient functions to form a gradient difference curve, and calculate the area under the curve to obtain: D(t) = {ΔB ik (t), ΔSEF1 ik (t), ΔMEF1 ik (t)} A(t) = ∫D(t)^2 * Δt Δ: Difference of each parameter between the left and right brains Δt: Time increment The numerical value A represents the left-right brain difference of the effective high-frequency and low-frequency signals retained after most of the interference is removed, and obtains the influencing component of the comprehensive change in the working state of neurons caused by local hypoxia. The numerical values G and A are normalized by: Cerebral hypoxia index (t) = (exp(G + A) / C) × 100 C: Matching constant t: Time series A quantitative real-time cerebral hypoxia index of 0 - 100 is obtained.

Images