Disclosure of Invention
The technical problem to be solved by the invention is to provide an AD nerve regulation and control system based on feature extraction and closed-loop ultrasonic stimulation so as to realize the diagnosis of AD of a specific cerebral cortex of a mouse and the detection of improvement conditions, thereby applying proper ultrasonic stimulation to the specific cerebral cortex area in real time.
In order to solve the technical problems, the invention adopts the following technical scheme:
The AD nerve regulation and control system based on characteristic extraction and closed-loop ultrasonic stimulation comprises a programmable ultrasonic signal generation module, an ultrasonic stimulation module, a signal acquisition and processing module, a closed-loop control module, a signal transmission and storage module, an upper computer and an electroencephalogram electrode, wherein the electroencephalogram electrode is connected to an experimental object, the programmable ultrasonic signal generation module sends a generated stimulation signal to the ultrasonic stimulation module, the ultrasonic stimulation module sends the stimulation signal to the experimental object in an ultrasonic stimulation mode, the signal acquisition and processing module is connected with the electroencephalogram electrode and is used for acquiring an electroencephalogram signal recorded by the electroencephalogram electrode, preprocessing and frequency division processing the electroencephalogram signal and sending the processed electroencephalogram signal to the closed-loop control module, the system utilizes a transcranial ultrasonic stimulation technology to stimulate a specific area of the cerebral cortex of the experimental object and implant the electroencephalogram electrode in the target area, then extracts time domain, nonlinear dynamics and multidimensional characteristics of different frequency signals, and the obtained multidimensional characteristics are used as input of a third-stage processor in the closed-loop control module, so that real-time detection is carried out, and the AD ultrasonic stimulation parameters are continuously regulated according to a class diagnosis result, and the weighting parameters of the experimental object are further restrained.
The technical scheme of the invention is further improved in that the experimental object selects a mouse, the ultrasonic stimulation module is arranged on the sports cortex of the mouse, and the brain electrode is implanted into the CA1 region of the hippocampus of the mouse.
The technical scheme of the invention is further improved by implanting a three-stage serial processor into the closed loop control module, wherein the three-stage serial processor is used for carrying out three-stage processing on the electroencephalogram signals after the frequency division processing of the signal acquisition and processing module, the first-stage processor is a strong classifier and carries out quick screening of suspected AD electroencephalogram signals, the screened electroencephalogram signals enter the second-stage processor, a multi-component modal decomposition algorithm is adopted in the second-stage processor to realize multi-channel input of signals, the time domain features and nonlinear dynamics features of the signals are extracted from the signal components obtained through decomposition, the signal components are combined at the same time, a new signal matrix is constructed, the spatial features of the signal matrix are extracted by adopting CSP, the three features obtained in the third-stage processor are combined, the multi-modal features of EEG signals are obtained, finally, the classification is carried out through SVM, if the classification result is not AD abnormal signals, otherwise, the calculation is stopped, and the diagnosis result is sent to the programmable ultrasonic signal generation module so as to achieve the aim of treatment by timely adjusting the stimulation parameters;
The signal transmission and storage module is used for receiving the working parameters of each module configured by the upper computer and the brain electrical signals transmitted by the closed-loop control module, and storing the brain electrical signals as a data set;
The upper computer is used for training parameters of the three-stage serial processor implanted in the closed-loop control module according to the data set, and communicating with the signal transmission and storage module in real time, continuously adjusting working parameters of each module in operation, updating various parameters of the three-stage serial processor implanted in the closed-loop control module, and displaying acquired electroencephalogram signals in real time;
The programmable ultrasonic signal generation module is used for changing the output of ultrasonic stimulation in real time according to the result obtained by the closed-loop control module or the instruction of the upper computer.
The technical scheme of the invention is further improved in that the first-stage processor is a strong classifier trained by adopting an Ada Boost algorithm.
The technical scheme of the invention is further improved in that the second-stage processor is used for extracting multi-mode characteristics, wherein the multi-mode data are from the disclosed AD mouse brain electrical data on one hand, and the mouse brain electrical data of a model group, a false stimulation group and a normal control group are selected on the other hand.
The technical scheme of the invention is further improved in that the components of a plurality of signals in the three-stage serial processor are obtained by adopting an MVMD method.
A further improvement of the solution according to the invention is that the step of preprocessing the EEG signal comprises filtering and noise reduction.
The technical scheme of the invention is further improved in that the SVM classifier can test the test set after training the training set to obtain the classification model.
An AD nerve regulation and control method based on feature extraction and closed loop ultrasonic stimulation comprises the following steps:
Step 1, respectively implanting the ultrasonic stimulation module and the electroencephalogram electrode into preset sites of a plurality of mice, wherein the preset sites are positioned in brain areas of the mice;
step 2, after all experimental mice implanted with the brain electrode recover for t time, regulating and controlling the output parameters of the ultrasonic stimulation signals of the ultrasonic stimulation module by using the programmable ultrasonic signal generating module;
step 3, after the ultrasonic transducer of the ultrasonic stimulation module receives the stimulation signal, performing ultrasonic stimulation on the intracranial brain preset site of the experimental mouse;
step 4, the signal acquisition and processing module is used for acquiring the electroencephalogram signals recorded on the electroencephalogram electrodes, the processed electroencephalogram signals are sent to the closed-loop control module, and whether the experimental mice need to adjust ultrasonic stimulation parameters or not is judged according to classification results through processing of the three-stage serial processors;
step5, the upper computer trains parameters of the three-stage serial processor implanted in the closed-loop control module according to the existing data set, and communicates with the signal transmission and storage module in real time;
step 6, according to the judgment result of the step 4, if the stimulation parameters do not need to be adjusted, the step 7 is carried out, otherwise, the step 3 is returned;
And 7, carrying out Morris water maze experiments on the experimental mice every T period to obtain evaluation indexes of Morris water maze experiments, wherein the evaluation indexes of Morris water maze experiments comprise escape latency and escape path length, judging the difference of the evaluation indexes of Morris water maze experiments under the interaction of groups and days, if the difference between the time of a third quadrant and the time of other quadrants among groups is larger than a preset threshold value, finishing regulation, otherwise, modifying ultrasonic stimulation parameters, and returning to the step 3.
The technical scheme of the invention is further improved in that the steps of Morris water maze experiment on the experimental mice comprise:
701, equally dividing a water maze into 4 areas, wherein the water maze is provided with a visible platform positioned above the water surface;
702, enabling the experimental mice to enter water, wherein the water inlet point is a pool wall at the middle point of each area;
703, acquiring a swimming track of the organism by using a CCD camera and storing the swimming track in a video acquisition card;
And 704, uploading the swimming track of the experimental mouse to a computer by the video acquisition card, carrying out image recognition on the swimming track of the experimental mouse by the computer to obtain an evaluation index of the Morris water maze experiment, wherein the escape latency is the time from entering water to boarding a visible platform of the experimental mouse.
By adopting the technical scheme, the invention has the following technical progress:
The AD nerve regulation and control system and method based on the feature extraction and the closed loop ultrasonic stimulation can detect whether the brain electrical signal of the specific cerebral cortex is an abnormal AD signal or not, and can adjust proper ultrasonic stimulation parameters according to the detection result, so that accurate monitoring and nerve regulation and control of the specific cerebral cortex can be realized.
According to the invention, the motor cortex of the mouse is selected as a stimulation target point, a transcranial ultrasonic stimulation paradigm is designed, a chronic stimulation experiment is carried out, an experimental control group is designed, and the transcranial ultrasonic stimulation effect is evaluated from the perspective of animal behaviours and brain electrical signals, so that the optimal stimulation parameters are found. In order to detect the nerve regulation effect, the difference of the ultrasonic stimulation of different parameters on AD nerve activity regulation is evaluated by recording the physiological signals acquired and analyzed by implanting brain electrodes into the hippocampus CA1 (AP: 2.06, ML: + -1.5, DV: 1.25) of the mice, and a thought is provided for the treatment parameter optimization of Alzheimer disease.
Detailed Description
It is noted that the terms "comprises" and "comprising," and any variations thereof, in the description and claims of the present invention and in the foregoing figures, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed or inherent to such process, method, article, or apparatus.
The invention is described in further detail below with reference to the attached drawings and examples:
As shown in fig. 1, the embodiment provides an AD nerve regulation and control system based on feature extraction and closed-loop ultrasonic stimulation, which comprises a programmable ultrasonic signal generation module 1, an ultrasonic stimulation module 2, an electroencephalogram electrode 3, a signal acquisition and processing module 4, a closed-loop control module 5, a signal transmission and storage module 6 and an upper computer 7;
the programmable ultrasonic signal generation module 1 can send the generated stimulation signal to the ultrasonic stimulation module 2 according to the processing result obtained by the closed-loop control module 5 or the instruction of the upper computer 7 to change the output of ultrasonic stimulation in real time, and the electroencephalogram electrode 3 is electrically connected to an experimental object (mouse) and is used for collecting the electroencephalogram signals of the mouse;
The signal acquisition and processing module 4 is electrically connected with the electroencephalogram electrode 3 and is used for acquiring an electroencephalogram signal recorded by the electroencephalogram electrode 3, performing serial processing such as analog-to-digital conversion, filtering and noise reduction on the acquired electroencephalogram signal and sending the processed electroencephalogram signal to the closed-loop control module 5.
The signal transmission and storage module 6 can transmit the received brain electrical signals to the upper computer 7 in a wired or wireless manner for real-time display and analysis of the brain electrical signals, and in a low-power consumption operation mode, the signal transmission and storage module 6 is not in physical connection with the upper computer 7, and directly stores the received brain electrical signals in an on-board SD memory for subsequent offline analysis and processing.
The closed-loop control module 5 performs multidimensional (time domain, nonlinear dynamics, airspace, etc.) feature extraction and classification on the electroencephalogram signal after receiving the preprocessed electroencephalogram signal, determines whether the AD aggravation occurs in the specific cerebral cortex in real time, and after the classification result is transmitted to the stimulus control sub-module 54, the stimulus control sub-module 54 configures different ultrasonic stimulus modes and parameters according to the configured stimulus parameters and the classification result and feeds back the different ultrasonic stimulus modes and parameters to the programmable ultrasonic signal generating module 1. The programmable ultrasonic signal generation module 1 applies corresponding ultrasonic stimulation pulses according to the received stimulation parameters to intervene on intracranial neuron conditions in corresponding areas, and the closed-loop intervention process of the system on AD diagnosis is completed.
Further, as shown in fig. 2, the closed-loop control module 5 includes a data receiving sub-module 51, a category diagnosis sub-module 52, a parameter configuration sub-module 53 and a stimulus control sub-module 54, where the data receiving sub-module 51 is used as an interface between the signal acquisition and processing module 4, the signal transmission and storage module 6 and the closed-loop control module 5, and is capable of being responsible for receiving and buffering the electroencephalogram signals of the signal acquisition and processing module 4 and configuring parameters of an upper computer transmitted by the signal transmission and storage module 6 through an SPI communication manner. After the data buffer in the data receiving sub-module 51 obtains a neural signal time sequence with a predetermined length, the neural signal sequence enters the category diagnosis sub-module 52, which is essentially a three-stage serial processor, to determine whether the current electroencephalogram signal is an AD signal emphasis segment. The classification result is transmitted to the stimulus control sub-module 54, and the stimulus control sub-module 54 transmits different stimulus mode parameters to the programmable ultrasonic signal generating module 1 according to the corresponding configuration parameters such as stimulus time, stimulus intensity, duty cycle and the like transmitted by the upper computer in the parameter configuration sub-module 53, and if the diagnosis result of the specific cerebral cortex nerve signal is an abnormal AD signal, the corresponding region is stimulated by corresponding ultrasonic.
Further, the classifier used by the closed loop control module 5 to construct the three-stage serial processor is obtained by Real AdaBoost algorithm based on the criteria that minimizes the loss function in the positive and negative sample sets in the training set. The classifier ci is composed of a threshold value and a segmented output function, and outputs one value when the corresponding characteristic value f of the signal is larger than the threshold value theta, or outputs another value otherwise. The piecewise function and the threshold value output by the classifier are obtained by training the acquired electroencephalogram signals.
The first stage processor is a strong classifier, and H (x) is obtained through training of a Real AdaBoost algorithm. The classifier ci used in the first stage corresponds to small calculated quantity features such as amplitude values, frequency spectrums and the like of different rhythms after frequency division, and is beneficial to quick screening of suspected abnormal AD signals.
H(x)=∑a=1,...na (2)
The second-stage processor further decomposes the signals subjected to rapid screening by adopting an MVMD method, extracts time domain features and nonlinear dynamic features of the signals from signal components obtained by decomposition, combines the signal components to construct a new signal matrix, and extracts spatial features of the signal matrix by adopting CSP.
The MVMD method realizes the change from single-channel input signals to multi-channel input signals, and can keep the frequency of each IMF component the same when decomposing data. The components obtained by decomposition are taken as the input of an iterator, the center frequency and the bandwidth are taken as the updating targets of the iterator, and the output of the iterator is the k components. Assuming that the signal of the C sampling channels is X (t), it can be expressed in mathematical form as [ X1(t),x2(t),…xC (t) ].
(1) Let k components be included in the signal first, and satisfy:
(2) In the vector uk (t), the data analysis is represented as Hilbert-Huang transform (HHT)And taking the same as a reference to obtain a single-sided frequency spectrum, and multiplying the single-sided frequency spectrum by an exponential termThe center frequency is adjusted. RecalculatingThe objective function is optimized to keep each obtained component as far as possible to form the original signal while minimizing the bandwidth of the component, and the following is the solved optimization problem:
Wherein, theIs an analytical expression form of the data.
(3) To solve this variation problem, a Lagrangian of the form:
(4) UpdatingAndFrom the updated values obtained, the magnitudes of uk (t) and the center frequency can be calculated, and thus the decomposed individual signal components can be obtained. The further update mode is:
The update frequency is:
After adopting HHT method, it can obtain the change characteristic of EEG signal in time direction by analyzing component characteristic, and obtain instantaneous energy H [ uk (t) ] according to the instantaneous amplitude of IMF, and can obtain information in frequency domain and amplitude change:
Uk(t)＝uk(t)+jH[uk(t)] (8)
Calculating an energy amplitude value for the sampled signal:
wherein n is the number of sampling points,Is the magnitude of the discrete signal i. The average instantaneous energy value reflects the change in the signal in the time domain and is denoted as F1.
Introducing information difference and analysis signal complexity of a multi-scale entropy observation signal in a plurality of modes, sampling the decomposed IMF function to obtain discrete signals of different modes, performing series analysis, average and dimension transformation, and finally obtaining a sample entropy value when the time sequence length is M:
SampEn(m,r,M)=-ln[Cm+1(r)/Cm(r)] (10)
The above calculations are repeated to obtain entropy features at multiple scales, which are combined to obtain multi-scale entropy features of the EEG signal, denoted F2.
The CSP spatial domain feature is that the IMF component is obtained and the sampling signals of the component are combined, the signal matrix is formed by the total number k of the component and the signal of the sampling point number n, namely k multiplied by n is taken as the object of CSP processing, the spatial domain feature of the CSP spatial domain feature is marked as F3 by taking the component of C3 and C4 as an example, and the matrix can be expressed as follows:
the third-stage processor combines the extracted multiple feature information, and the obtained multi-mode feature is denoted as f= { F1,F2,F3 }, and the whole processing procedure is shown in fig. 3. To avoid the difference in values of the different features, the extracted features are normalized:
Fe＝(Fe-μe)/σe,e=1,2,3 (12)
Wherein mue、σe represents the mean value and standard deviation, respectively, when the characteristic is e. And classifying the F which completes the normalized feature. Based on the common representation, an SVM classifier is introduced to obtain a final diagnosis result.
An AD nerve regulation and control method based on feature extraction and closed loop ultrasonic stimulation comprises the following steps:
To evaluate the therapeutic effect of transcranial ultrasound stimulation by control experiments, AD mice models were used and grouped by a method in which AD mice were randomized into 2 groups, including a stimulated group (ADT group), a sham stimulated group (ADs group), and healthy mice as normal control groups (WT group). The sham group was prepared by reserving a stimulation area in the motor cortex of AD mice and implanting the electroencephalogram electrode 3 in the hippocampus, but without applying any ultrasonic stimulation.
Step 1, respectively implanting the ultrasonic stimulation module 2 and the electroencephalogram electrode 3 into preset sites of mice in a stimulation group (ADT group), a pseudo stimulation group (ADS group) and a control group (WT group), wherein the preset sites are positioned in brain areas of the mice;
specifically, the mice in the experiment are placed in a gas anesthesia induction box, anesthesia is adjusted to 2.5L/min, and the mice are left for about 2 minutes until the toes of the mice are pinched without leg shrinking reaction. The chloral hydrate with the proportion of 1% is used for realizing surgical anesthesia by intraperitoneal injection according to the weight proportion.
Craniectomy was performed in the motor cortex (AP: -1.54, ML: + -1.5), forming a viewing window for the ultrasound stimulation and implanting a glass plate. The brain electric signal collecting electrode is implanted into the CA1 (AP: 2.06, ML: + -1.5, DV: 1.25) of the Hippocampus, the brain electric electrode for collecting/recording is implanted into the CA1 region of the mouse Hippocampus, and two skull nails are additionally arranged at the nasal bone position for grounding and reference.
Step 2, after all experimental mice implanted with the brain electrode 3 recover for t time, regulating and controlling the output parameters of the ultrasonic stimulation signals of the ultrasonic stimulation module 2 by using the programmable ultrasonic signal generation module 1;
Specifically, mice diagnosed with AD were treated after 1 week of recovery, at which time all mice were 5 months of age. The stimulus group mice receive the signals delivered by the stimulus control submodule 54 to the programmable ultrasound signal generating module 1, the programmable ultrasound signal generating module 1 sends the generated stimulus signals to the ultrasound stimulus module 2, and the ultrasound stimulus module 2 emits the stimulus signals in the form of ultrasound stimulus to the viewing window area of the implanted glass sheet.
Step 3, after the ultrasonic transducer of the ultrasonic stimulation module 2 receives the stimulation signal, performing ultrasonic stimulation on the intracranial brain preset site of the experimental mouse;
Step 4, the electroencephalogram signals recorded on the electroencephalogram electrodes 3 are collected by the signal collecting and processing module 4, the processed electroencephalogram signals are sent to the closed-loop control module 5, whether the experimental mice need to adjust ultrasonic stimulation parameters or not is judged according to classification results through processing of the three-stage serial processors, the signal transmitting and storing module 6 receives the electroencephalogram signals transmitted by the closed-loop control module 5, and the electroencephalogram signals are stored as a data set;
Step 5, the upper computer 7 trains parameters of the three-stage serial processor implanted in the closed-loop control module 5 according to the existing data set, and carries out real-time communication with the signal transmission and storage module 6;
step 6, according to the judgment result of the step 4, if the stimulation parameters do not need to be adjusted, the step 7 is carried out, otherwise, the step 3 is returned;
And 7, carrying out Morris water maze experiments on the experimental mice every T period to obtain evaluation indexes of Morris water maze experiments, wherein the evaluation indexes of Morris water maze experiments comprise escape latency and escape path length, judging the difference of the evaluation indexes of Morris water maze experiments under the interaction of groups and days, if the difference between the time of a third quadrant and the time of other quadrants among groups is larger than a preset threshold value, finishing regulation, otherwise, modifying ultrasonic stimulation parameters, and returning to the step 3.
Specifically, in order to evaluate the safety of ultrasonic stimulation, ensuring that anxiety-related side effects are not caused, a Morris water maze experiment was designed. The Morris water maze experiment is carried out on the experimental mice in each T period to obtain evaluation indexes of the Morris water maze experiment, namely, escape latency period and escape path length, whether the two indexes are obviously subject to the analysis of groups and days or not and obviously different in interaction of groups by days or not are obtained, and if obvious differences exist between the time of a third quadrant and the time of other quadrants among groups, the results show that the ADT group can effectively distinguish the quadrant where a platform is located from the other quadrants, and the distinguishing capacity of the other groups is inferior. Otherwise, the problem of the stimulation scheme is indicated, and the stimulation parameters can be adjusted.
The Morris water maze (Morris water maze, MWM) comprises a water maze, a computer, a video acquisition card, a CCD camera and other devices. Two virtual vertical lines are arranged in the pool to divide the pool into I, II, III, IV four quadrants uniformly, the water inlet point of the mouse is arranged on the pool wall at the midpoint of each quadrant, and a cylindrical visible platform with the diameter of 10cm is arranged at the middle of the third quadrant. Proper amount of compound coloring agent is added into water to be prepared into white, so that a behavior analysis system can track the swimming track of a mouse in the experimental process.
The Morris water maze test method for the experimental mice comprises the following steps:
1) Dividing the water maze into 4 areas, wherein the water maze is provided with a visible platform positioned above the water surface;
2) Enabling the experimental mice to enter water, wherein the water inlet point is a pool wall at the middle point of each area;
3) The swimming track of the experimental mouse is collected by using a CCD camera and is stored in a video acquisition card;
4) The video acquisition card uploads the swimming track of the experimental mouse to the computer, the computer performs image recognition on the swimming track of the experimental mouse to obtain an evaluation index of the Morris water maze experiment, and the escape latency is the time from entering water to boarding the platform of the experimental mouse.
In particular, the visible platform extends 1 cm above the water surface. Mice were placed in pools at different quadrant walls. The time it takes the mouse to find and board the visible platform within 60s was recorded as escape latency (ESCAPE LATENCY). If the mouse does not find the visible platform within 60s, the mouse is guided to the visible platform and placed on the visible platform for 15-20 s, and the escape latency period is recorded to be 60s. At the same time, the path length before the mouse escaped to the visible platform was recorded. If the escape latency and path length of each group were not significantly different, each group of mice was considered to have similar motor and visual abilities.
It should be noted that the above embodiments are merely for illustrating the technical solution of the present invention and not for limiting the same, and although the present invention has been described in detail with reference to the above embodiments, it should be understood by those skilled in the art that the technical solution described in the above embodiments may be modified or some or all of the technical features may be equivalently replaced, and these modifications or substitutions do not make the essence of the corresponding technical solution deviate from the scope of the technical solution of the embodiments of the present invention.