CN113208615B

CN113208615B - Continuous electroencephalogram monitoring and feedback system and method for cardio-pulmonary resuscitation instrument

Info

Publication number: CN113208615B
Application number: CN202110630934.3A
Authority: CN
Inventors: 李可; 张剑红; 徐峰; 陈玉国; 边圆; 潘畅; 王甲莉; 李贻斌; 蒋丽军; 徐凤阳
Original assignee: Shandong University; Qilu Hospital of Shandong University
Current assignee: Shandong University; Qilu Hospital of Shandong University
Priority date: 2021-06-07
Filing date: 2021-06-07
Publication date: 2023-02-28
Anticipated expiration: 2041-06-07
Also published as: CN113208615A

Abstract

The invention belongs to the field of electroencephalogram monitoring, and provides a continuous electroencephalogram monitoring and feedback system and method for a cardio-pulmonary resuscitation instrument. The system comprises an electroencephalogram signal monitoring module, a data processing module and a data processing module, wherein the electroencephalogram signal monitoring module is used for monitoring electroencephalogram signals of a patient in real time; the electroencephalogram signal analysis module is used for judging the compression quality based on the amplitude values of the upper boundary and the lower boundary of the real-time electroencephalogram signal and feeding the compression quality back to the cardio-pulmonary resuscitation apparatus so that the cardio-pulmonary resuscitation apparatus can repeatedly compress in a state of setting the compression quality; and the compression parameter monitoring module is used for judging the compression effect and feeding back the compression effect to the cardio-pulmonary resuscitation instrument based on the real-time compression parameters and comparing the compression parameters with a set threshold value so as to ensure the compression effect.

Description

Continuous electroencephalogram monitoring and feedback system and method for cardio-pulmonary resuscitation instrument

Technical Field

The invention belongs to the field of electroencephalogram monitoring, and particularly relates to a continuous electroencephalogram monitoring and feedback system and method for a cardio-pulmonary resuscitation instrument.

Background

The statements in this section merely provide background information related to the present disclosure and may not constitute prior art.

Brain damage is a common complication in the survivors of sudden cardiac arrest. After sudden cardiac arrest, cerebral tissue is not perfused sufficiently and is injured in the early stage, and 40 to 60 percent of unconscious patients cannot wake up after cardiopulmonary resuscitation; in severe cases, brain death, persistent vegetative state, etc. may occur. Most patients will begin irreversible brain damage within 4-6 minutes after cardiac arrest, and then transition to biological death over a few minutes. The time and degree of hypoxia before cardiac arrest, the time of starting cardiopulmonary resuscitation after cardiac arrest, the time of achieving cardiopulmonary resuscitation without spontaneous circulation, the time of insufficient circulation after cardiopulmonary resuscitation, and the like are all factors influencing the cerebral resuscitation effect. Therefore, the recovery of nerve function after cardiac arrest is an important consideration in assessing the quality of cardiopulmonary resuscitation.

Cardiopulmonary resuscitation is the most effective rescue method after cardiac arrest. Bare-handed resuscitation uses manual compression by the rescuer to restore the patient's spontaneous circulation, which is time consuming and laborious. The automatic compression of the patient by an automated device effectively solves this problem. However, the inventor finds that the existing cardio-pulmonary resuscitation apparatus lacks objective, continuous and accurate monitoring of brain function in the rescue process, and cannot adjust mechanical stimulation parameters of the cardio-pulmonary resuscitation apparatus according to the brain function state.

Disclosure of Invention

In order to solve the technical problems in the background art, the invention provides a continuous electroencephalogram monitoring and feedback system and method for a cardiopulmonary resuscitation apparatus, which can detect the brain function state of a patient with cardiac arrest in real time on line and dynamically adjust the mechanical stimulation parameters of the cardiopulmonary resuscitation apparatus.

In order to achieve the purpose, the invention adopts the following technical scheme:

in a first aspect of the invention, a continuous brain electrical monitoring and feedback system for a cardiopulmonary resuscitation machine is provided.

A continuous brain electrical monitoring and feedback system for a cardiopulmonary resuscitation instrument, comprising:

the electroencephalogram signal monitoring module is used for monitoring electroencephalogram signals of a patient in real time;

the electroencephalogram signal analysis module is used for judging the compression quality based on the amplitude values of the upper boundary and the lower boundary of the real-time electroencephalogram signal and feeding the compression quality back to the cardio-pulmonary resuscitation apparatus so that the cardio-pulmonary resuscitation apparatus can repeatedly compress in a state of setting the compression quality;

and the compression parameter monitoring module is used for judging the compression effect and feeding back the compression effect to the cardio-pulmonary resuscitation instrument based on the real-time compression parameters and comparing the compression parameters with a set threshold value so as to ensure the compression effect.

As an implementation manner, the electroencephalogram signal monitoring module includes an electrode, the electrode is configured to acquire an electroencephalogram signal and transmit the electroencephalogram signal to a radio frequency trap, the radio frequency trap is configured to filter an interference signal of the received electroencephalogram signal and transmit one of the filtered interference signal to an amplifier, and transmit the other of the filtered interference signal to an impedance network monitoring network to continuously monitor an impedance of the electrode, and meanwhile, the signal amplified by the amplifier is sequentially subjected to signal compression and a peak-to-peak detector to determine an amplitude value of an upper boundary and a lower boundary of the real-time electroencephalogram signal.

In one embodiment, in the peak-to-peak detector, a wavelet envelope analysis method is used to extract the wavelet envelope of the electroencephalogram signal.

In one embodiment, an LSTM classifier is used to classify the electroencephalogram signal in the electroencephalogram signal analysis module.

As an embodiment, the categories of the electroencephalogram signals are divided into four categories, which are:

continuous normal voltage: the lower boundary is 5-10 mu V, and the upper boundary is 10-50 mu V;

discontinuous voltage: the lower boundary is less than 5 mu V, and the upper boundary is more than 10 mu V;

burst suppression mode: the boundary fluctuation under the discontinuous background is 0-2 mu V and the burst amplitude is more than 25 mu V, and the boundary fluctuation is divided into BS + and BS-, wherein the BS + refers to the burst frequency of more than 100 times/h, and the BS-refers to the burst frequency of less than 100 times/h;

the low voltage, i.e. the brain wave is at rest, all amplitudes are below 5 μ V and close to 0.

In one embodiment, the electroencephalogram signal belongs to high-quality on-press when the category of the electroencephalogram signal is a continuous normal voltage or a discontinuous voltage.

As an embodiment, the compression parameters include compression depth, compression frequency and chest rebound.

As an embodiment, the compression parameter monitoring module comprises a three-axis acceleration sensor and a pressure sensor, wherein the three-axis acceleration sensor is used for acquiring compression depth and thoracic cavity rebound data when cardiopulmonary resuscitation is performed; the pressure sensor is used for collecting the compression frequency during the cardio-pulmonary resuscitation.

In the compression parameter monitoring module, when the compression parameters are compared with the set threshold value and all the compression parameters are normal, it is determined that the compression effect meets the set requirement; and when any one of the pressing parameters is abnormal, outputting warning information and determining that the pressing effect does not meet the set requirement.

A second aspect of the invention provides a method of operation of a continuous brain electrical monitoring and feedback system for a cardiopulmonary resuscitation machine.

A method of operation employing a continuous brain electrical monitoring and feedback system for a cardiopulmonary resuscitation apparatus as described above, comprising:

monitoring electroencephalogram signals of a patient in real time;

judging the compression quality based on the upper and lower boundary amplitude values of the real-time electroencephalogram signal and feeding back the compression quality to the cardio-pulmonary resuscitation apparatus so that the cardio-pulmonary resuscitation apparatus performs repeated compression in a state of setting the compression quality;

based on the real-time compression parameters and compared with the set threshold value, the compression effect is judged and fed back to the cardio-pulmonary resuscitation instrument so as to guarantee the compression effect.

Compared with the prior art, the invention has the beneficial effects that:

the invention provides a continuous electroencephalogram monitoring and feedback system for a cardio-pulmonary resuscitation (CPR) instrument, which adopts an electroencephalogram signal monitoring module to monitor electroencephalogram signals of a patient in real time; the electroencephalogram signal analysis module is used for judging the compression quality based on the amplitude values of the upper boundary and the lower boundary of the real-time electroencephalogram signal and feeding the compression quality back to the cardio-pulmonary resuscitation apparatus so that the cardio-pulmonary resuscitation apparatus can repeatedly compress the brain in the state of setting the compression quality; the compression parameter monitoring module is used for judging the compression effect and feeding back the compression effect to the cardiopulmonary resuscitation instrument based on the real-time compression parameters and comparing the compression parameter with a set threshold value so as to guarantee the compression effect, so that the brain function state of a patient suffering from cardiac arrest can be detected online in real time, and the mechanical stimulation parameters of the cardiopulmonary resuscitation instrument can be dynamically adjusted, so that the whole external chest compression closed-loop control system achieves the optimal resuscitation effect.

Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, are included to provide a further understanding of the invention, and are included to illustrate an exemplary embodiment of the invention and not to limit the invention.

FIG. 1 is a diagram of a cardiopulmonary resuscitation device of an embodiment of the present invention;

FIG. 2 is a schematic diagram of an electroencephalogram detector according to an embodiment of the present invention;

FIG. 3 is a flow chart of a quantitative analysis aEEG algorithm according to an embodiment of the present invention;

FIG. 4 is a graph of four aEEG activity patterns for a sudden cardiac arrest patient in accordance with an embodiment of the present invention;

FIG. 5 is a LSTM schematic diagram of an embodiment of the present invention;

FIG. 6 is a block diagram of an electroencephalogram signal classification algorithm according to an embodiment of the present invention;

the device comprises a heart-lung resuscitator, a brain electrical signal analysis module, a C3 electrode pasting site, a C4 electrode pasting site, a P3 electrode pasting site, a P4 electrode pasting site, a brain electrical signal real-time monitor, a continuous normal voltage, a continuous abnormal voltage, an explosion suppression mode and a brain electrical signal rest, wherein the heart-lung resuscitator 1 is an automatic chest compression cardiopulmonary resuscitator, the brain electrical signal analysis module 2 is a brain electrical signal analysis module, the C3 electrode pasting site 3 is a C4 electrode pasting site, the P3 electrode pasting site 5 is a P4 electrode pasting site, the brain electrical signal real-time monitor 7 is a brain electrical signal real-time monitor, the continuous normal voltage 8 is a continuous normal voltage, the continuous abnormal voltage is a continuous abnormal voltage, the explosion suppression mode is a burst suppression mode, and the brain electrical signal rest 11 is a brain electrical signal.

Detailed Description

The invention is further described with reference to the following figures and examples.

It is to be understood that the following detailed description is exemplary and is intended to provide further explanation of the invention as claimed. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.

It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the invention. As used herein, the singular forms "a", "an", and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.

In the present invention, terms such as "upper", "lower", "left", "right", "front", "rear", "vertical", "horizontal", "side", "bottom", and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the drawings, and are only terms of relationships determined for convenience of describing structural relationships of the parts or elements of the present invention, and are not intended to refer to any parts or elements of the present invention, and are not to be construed as limiting the present invention.

In the present invention, terms such as "fixedly connected", "connected", and the like are to be understood in a broad sense, and mean either a fixed connection or an integrally connected or detachable connection; may be directly connected or indirectly connected through an intermediate. The specific meanings of the above terms in the present invention can be determined according to specific situations by persons skilled in the art, and should not be construed as limiting the present invention.

From a continuous electroencephalogram monitoring perspective, although the Glasgow Coma Score (GCS) is the most common clinical indicator of brain injury, it has been reported that the accuracy of predicting post-Coma results is lower than that of electroencephalography (EEG). EEG has now proven valuable in assessing survival probability. EEG performance reflects the brain recovery process and is the most common tool for predicting the prognosis of cardiac arrest outside of clinical examinations. The continuous electroencephalogram mode can provide important information of brain functions of a coma patient, is expected to objectively and accurately evaluate the nerve function recovery condition of a cardiac arrest patient, and brings a new method for cardio-pulmonary resuscitation real-time brain function monitoring. However, there is currently no technical implementation of continuous brain function monitoring for cardiopulmonary resuscitation instruments.

As shown in fig. 1, the continuous electroencephalogram monitoring and feedback system for a cardiopulmonary resuscitation apparatus of the present embodiment includes an electroencephalogram signal monitoring module, an electroencephalogram signal analyzing module, and a compression parameter monitoring module.

In the embodiment, the electroencephalogram signal monitoring module is implemented by an electroencephalogram real-time monitor 7, and is used for monitoring electroencephalogram signals of a patient in real time.

The time and degree of hypoxia before cardiac arrest, the time of starting cardiopulmonary resuscitation after cardiac arrest, the time of realizing cardiopulmonary resuscitation without spontaneous circulation, the time of insufficient circulation after cardiopulmonary resuscitation, and the like are all factors influencing the cerebral resuscitation effect, and in order to further clarify the recovery condition of the nerve function of a patient during cardiopulmonary resuscitation, the EEG is selected as the primary method for investigation and confirmation.

Conventional approaches to acquiring a patient's EEG signal have limitations because they require continuous operation by highly skilled personnel and involve complex and cumbersome equipment, and thus EEG is only suitable for intermittent use. Due to cardiac arrest, it is difficult for conventional electroencephalograms to identify a temporary critical state, and without diagnosis, a patient may experience a condition of worsening, increased brain damage, and the like. Brain damage may be persistent during cardiopulmonary resuscitation, which is difficult to assess in comatose patients. The difficulty in evaluating and transmitting large amounts of data requires a reduction technique that enables observation of features of interest, such as critical patterns. Currently, the traditional method of reducing the amount of electroencephalogram data is Amplitude-integrated EEG (aeg), which originates from a brain function monitor (CFM).

Conventional techniques for recording electroencephalograms produce large amounts of data. To address this problem, CFM technology has been proposed to display in real time whether brain activity is abrupt, slowing down or sustained during cardiopulmonary resuscitation, whose basic content is EEG, an EEG monitor is a method of compressing the EEG signal for long periods of time, which can be used to monitor brain function in intensive care settings, as well as for neurological prognosis in cardiac arrest patients, by placing electrodes on the patient's scalp, producing a trace of electrical activity, which is then displayed on a semi-logarithmic circle of time-varying peak-to-peak amplitude, using a reduced number of electrodes to generate a single channel (two electrodes) or dual channel (four electrodes) EEG trace, using an algorithm to modify and compress the EEG signal compared to conventional EEG recordings. In brief, the aag technique amplifies, filters, and compresses the raw electroencephalogram signals from the dual top electrodes using the continuous electroencephalogram amplitude of a single channel to obtain a simplified electroencephalogram waveform, thereby enabling assessment of the trend of cortical brain activity.

As a specific implementation manner, the electroencephalogram signal monitoring module includes an electrode, the electrode is used for collecting an electroencephalogram signal and transmitting the electroencephalogram signal to a radio frequency trap, the radio frequency trap is used for filtering an interference signal of the received electroencephalogram signal and transmitting one path of the filtered interference signal to an amplifier, and the other path of the filtered interference signal is transmitted to an impedance network monitoring network to continuously monitor an electrode impedance, and meanwhile, the signal amplified by the amplifier is sequentially subjected to signal compression and a peak-to-peak detector to determine an amplitude value of an upper boundary and a lower boundary of a real-time electroencephalogram signal, as shown in fig. 2.

Wherein the impedance monitoring network injects a high frequency signal between the electrodes through a 70M omega resistor and a small isolation capacitor. The change in electrode impedance and radio frequency capacitance causes a phase change in the signal relative to the source, which is extracted from the electrical signal after passing through a parametric amplifier and compared to the source signal phase. The internal logic of this network is connected as follows: for zero impedance, the output is zero, then the impedance is increased and the output is a positive move.

In this embodiment, two channels, that is, four electrodes are used to acquire electroencephalogram signals, for example, the C3 electrode, the C4 electrode, the P3 electrode and the P4 electrode in fig. 1, where 3 in fig. 1 is a C3 electrode pasting site, 4 is a C4 electrode pasting site, 5 is a P3 electrode pasting site, and 6 is a P4 electrode pasting site.

The specific working principle of the electroencephalogram signal monitoring module is as follows:

the signals from a pair of electrodes are amplified and passed through a wide frequency filter. The signal is then compressed by a logarithmic amplifier and enters a peak rectifier, which has a short time constant, so that the EEG signal will vary with short-term fluctuations in the amplitude of the activity, and since the recording electrode remains in place for a long time, it is necessary to monitor its conductivity state in some way. To this end, a special network is added in parallel with part of the brain function monitoring system to continuously monitor the electrode impedance. The output of the impedance monitor is input into a separate trace of the chart recorder. Thus, when the electrode contact begins to deteriorate, a preliminary warning of the need for replacement is received; similarly, when electrode artifacts or amplifier failure occur, an indication is also given.

In fig. 2, the rf trap is located on a parametric amplifier with an input impedance of 10M Ω and a common mode rejection ratio of greater than 110dB. It is cut off below 2cps to suppress factors such as low frequency fluctuation caused by perspiration, and cut off above 15cps, and has a high cut-off rate even above 30 cps. The signal is then passed through a 2-15Hz bandpass asymmetric filter having a slope that is approximately the inverse of the normal EEG, which acts to bring the EEG within the passband to more stably measure the non-sinusoidal components. After filtering, the signal will pass through a waveform amplitude compression network. After peak-to-peak rectification, the output signal is effectively a smooth line that passes through the peak of the compressed signal, leaving a dc component.

The electroencephalogram signal analysis module 2 is used for judging the compression quality based on the amplitude values of the upper and lower boundaries of the real-time electroencephalogram signal and feeding the compression quality back to the cardio-pulmonary resuscitation apparatus, so that the cardio-pulmonary resuscitation apparatus can repeatedly compress the brain in the state of setting the compression quality. The cardiopulmonary resuscitation device in this embodiment is an automated chest compression cardiopulmonary resuscitation device 1.

This example classifies the aagc background pattern produced by a comatose patient following cardiac arrest while undergoing cardiopulmonary resuscitation: (1) Continuous normal voltage is 8, the lower boundary is 5-10 mu V, and the upper boundary is 10-50 mu V; (2) discontinuous voltage 9 (slight anomaly): the lower boundary is less than 5 mu V, and the upper boundary is more than 10 mu V; (3) burst suppression mode 10 (severe abnormality): the boundary fluctuation under the discontinuous background is 0-2 muV, the explosion amplitude is more than 25 muV, and the boundary fluctuation is divided into BS + (the explosion frequency is more than 100 times/h) and BS- (the explosion frequency is less than 100 times/h); (4) The low voltage, brain electrical rest 11, is sustained, all amplitudes below 5 μ V are close to 0, as shown in FIG. 4.

In the above classification methods, the main focus is to calculate the upper and lower boundary amplitudes of the aagc, which determine the top and bottom envelopes of the trace, reflecting the peak-to-peak amplitude of the EEG signal. Unfortunately, as an important indicator of discrimination, the amplitude of the aagc border is typically measured semi-subjectively by the naked eye. In order to quantitatively evaluate dynamic electroencephalogram data and to generalize the dynamic electroencephalogram method in a wider field of application, the present embodiment introduces a detailed quantitative EEG signal processing method to obtain a compact dynamic electroencephalogram with well-defined upper and lower boundaries.

Specifically, an asymmetric filter is integrated in the CFM to provide higher frequencies with greater amplification. Based on the EEG characteristics, the present example produced a flat band-pass filter and given it an asymmetric gain with a slope of 12dB (a linear phase FIR filter designed using Parks-McClellan algorithm) over a frequency range of 2-15 Hz. The filter performs weighted amplification on components of different frequencies in the electroencephalogram while suppressing low-frequency artifacts. Absolute value evaluation is employed in digital algorithms, replacing traditional electronic components to obtain the rectified brain electrical signal.

The purpose of the aag method is to monitor brain function by displaying amplitude trends in brain activity. What characterizes the trend of amplitude variation is the boundary (i.e., envelope) of the electroencephalogram waveform, in CFM (brain function monitoring), conventional envelope detection is implemented by using a diode and an impedance monitor, except that the present embodiment uses a five-order butterworth filter with a gain of 2 to rectify and low-pass filter the signal to obtain the electroencephalogram envelope signal. Finally, to obtain a compact aeg top and bottom boundary, we reduce the entire waveform to a series of vertical lines, time and amplitude compressing the envelope using non-overlapping epochs of 15 seconds duration. More specifically, the corrected envelope of the brain electrical signal is divided into non-overlapping time segments of 15 seconds duration, and the maximum and minimum amplitudes of each epoch are detected as the upper and lower endpoints of the associated aEEG vertical line.

The upper and lower boundaries of the aag reflect the maximum and minimum fluctuations in brain activity. In the aag tracking consisting of vertical lines, the upper and lower boundaries are easily delineated using connecting lines of end points. To obtain a smooth and representative margin, the median amplitude for each consecutive 20 endpoints is defined in the algorithm as the upper and lower bounds of the aagc. The signal flow diagram of the algorithm is shown in fig. 3.

The present embodiment uses the amplitude values of the upper/lower boundaries of each envelope as the aag feature to characterize the patient's neurological recovery during cardiopulmonary resuscitation.

In this embodiment, a new method for classifying EEG using wavelet packet analysis and LSTM classifier is proposed for accurate classification of EEG features collected in real time. Firstly, discrete wavelet transform is carried out on the electroencephalogram signals to obtain basic information of amplitude modulation and frequency modulation. Then, in order to extract the wavelet envelope, a predetermined wavelet sub-band is selected and subjected to Hilbert Transform (HT). And finally, identifying the EEG signal by using an LSTM classifier consisting of an input layer, an LSTM hidden layer and an output layer, wherein the specific process is as follows:

transforming by Discrete Wavelet (DWT):

in the formula (2), the reaction mixture is,

for wavelet series,. Psi. (t) is a wavelet basis function satisfying a certain condition, j and k represent frequency resolution and amount of time shift, respectively, f _j (t) represents the component of the signal f (t) at a certain scale. In practical application, the Mallat algorithm is used to perform finite layer decomposition on the signal, and the following results are obtained:

wherein L is the number of decomposition layers, A _L For a low-pass approximation component, D _j Are detail components at different scales. Thereby, the entire frequency band of the signal is divided into several sub-bands. Discrete wavelet transforms analyze signals in different frequency bands and different resolutions by decomposing the signals into coarse approximations and detailed information. In the multi-resolution decomposition of the signal x (t) in the program, each stage consists of two digital filters and two down-samplers. The downsampled outputs of the first high-pass and low-pass filters provide the detailed D1 and the approximate A1, respectively. The first approximation A1 is further decomposed and the process continues.

To extract the envelope of each subband signal, we perform HT on each subband signal. HT is a classical method of extracting the envelope signal. The method can effectively extract the envelope of the narrow-band carrier signal, and is widely applied to the envelope extraction of the electroencephalogram signal. After HT, the signal amplitude is unchanged and the phase is changed. The negative and positive frequencies are +90 and +90 phase shifts, respectively. post-HT signal

Comprises the following steps:

wherein x (t) is a signal before HT.

Thus, the analytical signal is obtained as:

the amplitude of g (t) is the envelope of the original signal:

specifically, the compression parameter monitoring module is used for judging the compression effect and feeding back the compression effect to the cardiopulmonary resuscitation instrument based on real-time compression parameters and comparing the compression parameters with a set threshold value so as to guarantee the compression effect.

The LSTM model, as in fig. 5, introduces a new structure called a memory unit, LSTM designs a memory unit specifically to hold history information, and the updating and utilization of the history information is controlled by 3 gates: input gate, forget to remember gate, output gate.

Long-Short Term Memory neural networks (LSTMs) are one type of RNN. The original RNN is trained, and as the training time is longer and the number of network layers is increased, the RNN neural network cannot use earlier information. That is, the farther the sequence is input, the smaller the influence of correct weight change, and the problem of gradient explosion or gradient disappearance easily occurs, so that the longer sequence data cannot be processed, and information of long-distance data cannot be acquired. The LSTM changes the influence mode of historical data on the current hidden layer by adding an input gate, an output gate and a forgetting gate on the hidden layer, effectively solves the problem that the RNN is difficult to solve and artificially prolongs the time task, and solves the problem that the RNN is easy to have gradient disappearance.

LSTM is implemented by the following complex function:

Z _t ＝g(W _z x _t +R _z y _t-1 +b _z ) (7)

i _t ＝σ(W _i x _t +R _i y _t-1 +P _i ⊙c _t-1 +b _i ) (8)

f _t ＝σ(W _f x _t +R _f y _t-1 +P _f ⊙c _t-1 +b _f ) (9)

c _t ＝z _t ⊙i _t +c _t-1 ⊙f _t (10)

o(t)＝σ(W _o x _t +R _o y _t-1 +P _o ⊙c _t-1 +b _o ) (11)

y(t)＝h(c _t )⊙o _t (12)

where σ, g, and h are point-wise nonlinear activation functions. Logic

For gate activation functions, hyperbolic tangent function (g (x) = h (x) = tanh (x)) is generally used as block input and output activation functions, and a point-by-point multiplication of two vectors indicates as [, ] W _Z ，W _i ，W _f ，W _o "is input weight," R _z ，R _i ，R _f ，R _o "is a recursive layer weight," P _z ，P _i ，P _f ，P _o "is the weight of peephole" b _z ，b _i ，b _f ，b _o "is the offset weight," Zt "is the block input function in LSTM," i _t "is an input gate," f _t "as forget gate," o (t) "is output gate function," c _t "are neuron functions in four neural network layers (fully connected layers) inside the LSTM," y (t) "are block output functions in the LSTM.

A single sub-band signature sequence contains too little frequency domain information. The present embodiment utilizes a combination of multiple sub-band signature sequences to provide different frequency domain information. However, since the time series of each subband envelope is not uniform, it is not possible to simultaneously input a plurality of subband time series to the LSTM classifier and vectorize the combination of the plurality of subband time series. Therefore, a sliding rectangular window is used to obtain a fixed-length time series, and the statistical feature combinations of multiple subbands over each window are computed as the features for each window. The feature extraction process is as follows:

(1) A rectangular window using overlapping windows is slid over the original EEG. Each window contains a separate EEG segment.

(2) Each segment of the brain electrical signal within the sliding window is decomposed into different frequency bands using discrete wavelet transform.

(3) And carrying out heat treatment on the sub-band subjected to discrete wavelet transform decomposition to obtain a sub-band envelope.

(4) And characterizing the envelope spectrum of the electroencephalogram signal by utilizing the statistical characteristics.

And decomposing each electroencephalogram fragment in a channel sliding window by using discrete wavelet transform, wherein the size of a rectangular window is 1s. And (3) carrying out 5-level decomposition on the electroencephalogram signals by using a Daubechies4-tap wavelet. Thereafter, the DWT decomposed subbands are HT-processed to obtain subband envelopes. And combining all wavelet envelope statistical characteristics extracted from the sub-bands to obtain a multi-dimensional characteristic vector as the characteristic of each electroencephalogram time sequence. And finally, taking the formed multidimensional characteristic time sequence as the input of the LSTM neural network.

The classification process can be divided into two parts: the first part is to optimize the classifier based on a training data set. Inputting the wavelet envelope characteristics of the training set into an LSTM classifier, and training the whole network through supervised learning. In the training process, a full-back-propagation (full-BPTT) algorithm is adopted to transfer the error from the upper layer to the lower layer, minimize the error and update the network parameter set. The second part is the analysis of the test data set. And inputting the wavelet envelope characteristics of the test set into the trained LSTM classifier, and evaluating the performance of the LSTM classifier according to the classification result. And then, combining the single classifiers into one classifier based on an additive model by using an algorithm such as AdaBoost, and training the network parameters and the weight of each classifier.

In the embodiment, DWT and HT are adopted to extract wavelet envelope characteristics of EEG signals, and an LSTM recurrent neural network is used as a classifier to classify and identify EEG signals. The algorithm framework is as in figure 6.

In a specific embodiment, the Cerebral Blood Flow (CBF) drops below the level required to maintain metabolic activity of brain cells before cardiac arrest, and then CBF is completely stopped within a few seconds. Specifically, if the CBF drops from normal flow 50ml/100g min to 25-35ml/100g min during ischemia in brain tissue, the rhythm of the cortical high frequency electrical activity (β -wave, α -wave) of the four layers and VOF under electroencephalographic monitoring will be affected, e.g., the ischemia reaches the critical ischemia threshold of 17-18ml/100g min, and the slow frequencies (θ and δ) produced by the cortical cells of the thalamus and II-VI layers will gradually increase. The initiation of cardiopulmonary resuscitation often fails to meet the metabolic needs of the brain, which is clinically manifested as brain dysfunction such as brainstem reflex impairment and loss of consciousness. Based on the fact that the CBF loss is initially shown on the electroencephalogram as an isoelectric line which progresses within 2-20s, namely the electroencephalogram resting mode mentioned above, when the artificial chest compression is performed, the compression depth or frequency declines with time, the present embodiment provides a closed-loop automatic chest compression cardiopulmonary resuscitation optimization control system based on the chest compression resuscitation theory and the intelligent control theory, and screens out the aag four types of background activity modes for quantitative analysis to provide an effective real-time feedback technology for cardiopulmonary resuscitation, so as to improve the cardiopulmonary resuscitation quality index performed during the cardiac arrest. During cardiopulmonary resuscitation, the normal electroencephalogram signals are originally isoelectroformed traces, and then an outbreak suppression mode similar to a variable duration appears, which indicates that the cerebral metabolic rate of a patient is sharply reduced at the moment, and the patient is in the outbreak suppression mode for a long time, which indicates that the patient is in a dangerous physical sign state and is easy to cause brain death.

The embodiment carries out real-time quality monitoring and follow-up medical instruction by acquiring real-time physiological parameters of a patient, and the parameters needing to be acquired are divided into compression parameters and physiological parameters. Wherein the pressing parameters include: compression depth, compression frequency, thoracic rebound. Physiological parameters: aagG.

Specifically, the cardiopulmonary resuscitation quality index is as follows: the pressing frequency is as follows: suggesting 10-120cpm; pressing depth: 5-6CM; rebound of the thoracic cavity: maintaining sufficient thoracic recoil. Judging the resuscitation effect through the amplitude values of the upper and lower boundaries of the aEEG: aEEG: in a continuous low pressure mode, no resuscitation signs; after automatic chest compression cardiopulmonary resuscitation, the aag transitions from a sustained low pressure mode to a brief burst suppression mode, indicating low quality compression; aaeeg is in discontinuous normal voltage mode, representing high quality compressions; the patient's aagc is at a continuous normal voltage and the patient resumes spontaneous circulation.

Therefore, in the automatic chest compression process, two extreme dangerous situations, such as "sustained low voltage" and "burst suppression mode", should be avoided as much as possible, when the aagc presents the above dangerous situation, the automatic cardiopulmonary resuscitation apparatus will give a warning to prompt improper compression, specifically, the cardiopulmonary resuscitation flow state diagram is shown in fig. 6, four compression parameters and physiological parameters are adopted to monitor the CPR quality, including: compression depth, compression frequency, aagc and chest rebound. The status of CPR will give an early warning during CPR as long as one parameter does not meet the best medical practice. The input equipment of the automatic cardio-pulmonary resuscitation instrument comprises a physiological sensor and an operation data acquisition sensor, wherein a three-axis acceleration sensor is used for acquiring compression depth and thoracic cavity rebound data during cardio-pulmonary resuscitation; pressure sensors are used to acquire the compression frequency during cardiopulmonary resuscitation and brain electrodes are used to acquire the patient's eeg pattern. Cardiopulmonary resuscitation is performed by chest compression of the sternum until spontaneous circulation is restored. In the embodiment, based on feedback monitoring of the compression parameters and the physiological parameters, advantages and disadvantages need to be fully balanced in the chest compression process, so that the chest compression effect with the highest cost performance is obtained.

In other embodiments, the present embodiment needs to automatically adjust the compression depth, compression frequency, etc. of the chest compression device according to the warning indication given by the resuscitation apparatus (e.g. insufficient compression depth, insufficient frequency, insufficient chest rebound, unfavorable aEEG amplitude value, etc.), and automatically perform chest compression according to the target compression index value.

To maximize the robustness and adaptability of closed-loop automatic chest compression control, for example: and selecting a fuzzy PID controller to carry out automatic closed-loop control and adjustment on the pressing depth. The fuzzy PID controller is utilized to integrate the advantages of the traditional PID control and the fuzzy control, the successful practical experience of the chest compression operation of emergency personnel is fully utilized, the PID control parameters of the automatic chest compression device are automatically adjusted according to the actual situation on site, the excellent control function of the PID controller is fully exerted, and the whole closed-loop chest compression control system achieves the optimal resuscitation effect.

The working method of the continuous electroencephalogram monitoring and feedback system for the cardiopulmonary resuscitation instrument comprises the following steps:

monitoring electroencephalogram signals of a patient in real time;

The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A continuous electroencephalogram monitoring and feedback system for a cardiopulmonary resuscitation instrument, comprising:

the electroencephalogram signal analysis module is used for judging the compression quality based on the amplitude values of the upper boundary and the lower boundary of the real-time electroencephalogram signal and feeding the compression quality back to the cardio-pulmonary resuscitation apparatus so that the cardio-pulmonary resuscitation apparatus can repeatedly compress in a state of setting the compression quality; in an electroencephalogram signal analysis module, classifying the electroencephalogram signals by adopting an LSTM classifier;

the compression parameter monitoring module is used for judging the compression effect based on the real-time compression parameters and comparing the real-time compression parameters with a set threshold value, and feeding the compression effect back to the cardiopulmonary resuscitation instrument to ensure the compression effect;

the electroencephalogram signal monitoring module comprises an electrode, the electrode is used for collecting electroencephalogram signals and transmitting the electroencephalogram signals to a radio frequency wave trap, the radio frequency wave trap is used for filtering interference signals of the received electroencephalogram signals and then transmitting one path of interference signals to an amplifier, the other path of interference signals is transmitted to an impedance network monitoring network to continuously monitor the impedance of the electrode, and meanwhile, the signals amplified by the amplifier are sequentially subjected to signal compression and a peak-to-peak detector to determine the amplitude values of the upper boundary and the lower boundary of the real-time electroencephalogram signals; wherein, in the peak-to-peak detector, a wavelet packet analysis method is adopted to extract a wavelet packet envelope of the electroencephalogram signal;

wherein, the EEG is classified by wavelet packet analysis and LSTM classifier, and the process comprises:

firstly, carrying out discrete wavelet transformation on an electroencephalogram signal to acquire basic information of amplitude modulation and frequency modulation; then, in order to extract the wavelet envelope, selecting a preset wavelet sub-band to perform Hilbert transform; and finally, identifying the EEG signal by using an LSTM classifier consisting of an input layer, an LSTM hidden layer and an output layer, wherein the specific process is as follows:

by discrete wavelet transform:

in the formula (2), phi _j,k ＝2 ^-j/2 ψ(2 ^-j/2 t-k) is a wavelet sequence, ψ (t) is a wavelet basis function satisfying a certain condition, j and k respectively represent frequency resolution and time shift amount, f _j (t) indicates that the signal f (t) is at a certain scaleA component of (a);

adopting a sliding rectangular window to obtain a time sequence with a fixed length, and calculating a statistical characteristic combination of a plurality of sub-bands on each window as the characteristic of each window; the feature extraction process is as follows:

(1) Sliding over the original EEG using rectangular windows of overlapping windows, each window containing a separate EEG segment;

(2) Decomposing each section of the brain electric signals in the sliding window into different frequency bands by utilizing discrete wavelet transformation;

(3) Carrying out heat treatment on the sub-band subjected to discrete wavelet transform decomposition to obtain sub-band envelope;

(4) Representing the envelope spectrum of the electroencephalogram signal by utilizing the statistical characteristics;

decomposing each electroencephalogram segment in a channel sliding window by using Discrete Wavelet Transform (DWT), carrying out Hilbert Transform (HT) on Discrete Wavelet Transform (DWT) decomposition sub-bands to obtain sub-band envelopes, and combining all wavelet envelope statistical characteristics extracted from the sub-bands to obtain a multi-dimensional characteristic vector as the characteristic of each electroencephalogram time sequence; and finally, taking the formed multidimensional characteristic time sequence as the input of the LSTM neural network.

2. The continuous electroencephalogram monitoring and feedback system for a cardiopulmonary resuscitation instrument of claim 1, wherein the categories of electroencephalogram signals are classified into four categories, respectively:

burst suppression mode: the boundary fluctuation under discontinuous background is 0-2 muV and the burst amplitude is more than 25 muV, and the boundary fluctuation is divided into BS + and BS-, wherein the BS + is the burst frequency more than 100 times/h, and the BS-is the burst frequency less than 100 times/h;

the low voltage, resting of the brain, is sustained, all amplitudes are below 5 μ V and close to 0.

3. The continuous electroencephalogram monitoring and feedback system for cardiopulmonary resuscitation devices of claim 2, wherein the category of electroencephalogram signals is of high-quality on-press when the category is continuous normal voltage or discontinuous voltage.

4. The continuous brain electrical monitoring and feedback system for cardiopulmonary resuscitation device of claim 1, wherein said compression parameters comprise compression depth, compression frequency, and chest rebound.

5. The continuous brain electrical monitoring and feedback system for a cardiopulmonary resuscitation device of claim 4, wherein said compression parameter monitoring module comprises a three-axis acceleration sensor and a pressure sensor, said three-axis acceleration sensor for acquiring compression depth and thorax rebound data during cardiopulmonary resuscitation; the pressure sensor is used to acquire the compression frequency during cardiopulmonary resuscitation.

6. The continuous brain electrical monitoring and feedback system for cardiopulmonary resuscitation device of claim 1, wherein in the compression parameter monitoring module, when the compression parameters are compared with a set threshold value and all the compression parameters are normal, it is determined that the compression effect meets the set requirement; and when any one of the pressing parameters is abnormal, outputting warning information and determining that the pressing effect does not meet the set requirement.

7. A method of operation employing the continuous brain electrical monitoring and feedback system for a cardiopulmonary resuscitation machine of any one of claims 1-6, comprising:

monitoring electroencephalographic signals of a patient in real time;

judging the compression quality based on the amplitude values of the upper and lower boundaries of the real-time electroencephalogram signal and feeding back the compression quality to the cardio-pulmonary resuscitation apparatus so that the cardio-pulmonary resuscitation apparatus performs repeated compression in a state of setting the compression quality;

based on the real-time compression parameters, comparing the real-time compression parameters with a set threshold value to judge the compression effect and feeding the compression effect back to the cardiopulmonary resuscitation instrument to ensure the compression effect;

classifying EEG by wavelet packet analysis and LSTM classifier, wherein wavelet packet analysis method is adopted to extract wavelet envelope of EEG signal; classifying the electroencephalogram signals by adopting an LSTM classifier; the process of classifying EEG includes:

by discrete wavelet transform:

in the formula (2), phi _j,k ＝2 ^-j/2 ψ(2 ^-j/2 t-k) is a wavelet sequence, ψ (t) is a wavelet basis function satisfying a certain condition, j and k represent frequency resolution and a time shift amount, respectively, f _j (t) represents the component of the signal f (t) at a certain scale;

(2) Decomposing each segment of the brain electric signals in the sliding window into different frequency bands by utilizing discrete wavelet transform;

(4) Representing the envelope spectrum of the electroencephalogram signal by utilizing statistical characteristics;

decomposing each electroencephalogram segment in a channel sliding window by utilizing discrete wavelet transform, performing Hilbert Transform (HT) on a Discrete Wavelet Transform (DWT) decomposition sub-band to obtain a sub-band envelope, and combining all wavelet envelope statistical characteristics extracted from the sub-band to obtain a multi-dimensional characteristic vector as the characteristic of each electroencephalogram time sequence; and finally, taking the formed multidimensional characteristic time sequence as the input of the LSTM neural network.