On-chip brain photostimulation system and signal processing method

Abstract

The invention discloses an on-chip brain photostimulation system and a signal processing method. According to the system, a light beam emitted by a light source module is collimated through a collimation module, an incident light beam is transmitted to a digital micromirror device through a collimation system, and a patterned neural optical genetic stimulation image is generated by controlling deflection of a plurality of micromirror units. The method comprises the following steps of: acquiring an image, projecting the image onto an on-chip brain through a microscope light path to carry out optical genetic stimulation, realizing real-time observation by utilizing an image acquisition module, recording two-dimensional on-chip brain nerve electric signals through electrodes of a multi-electrode array, and carrying out time domain, frequency domain and kinetic analysis on the signals recorded in real time. And real-time synchronization of optical genetic stimulation and electric signal recording of the two-dimensional on-chip brain is realized.

Description

Technical Field

[0001] The present invention relates to the fields of neuroscience and optogenetic technology, and more particularly to a on-chip brain light stimulation system and signal processing method. Background Art

[0002] The brain is the most complex and important organ of the human body. The research and regulation of brain nerve activities are of extremely important significance in the fields of neuroscience, medicine, etc. As an emerging means of neural regulation, optogenetic technology for the brain precisely stimulates brain neurons through light signals, providing a new approach for studying brain functions and treating nervous system diseases.

[0003] With the continuous development of micro-nano technology, the application of on-chip systems in the biomedical field has become increasingly widespread. Combining optogenetic technology for the brain with on-chip systems to achieve two-dimensional on-chip brain light stimulation and signal processing can improve the accuracy and efficiency of optogenetic technology for the brain, and at the same time realize the real-time acquisition and processing of brain nerve electrical signals, providing a more powerful tool for in-depth research on the mechanisms of brain nerve activities.

[0004] Traditional optogenetic devices for the brain usually adopt macroscopic light sources and optical path systems, making it difficult to achieve high-resolution and precise stimulation of neuron populations inside the brain. The stimulation modes and parameter adjustments of existing optogenetic systems are relatively fixed, and it is difficult to make real-time and flexible adjustments according to different experimental requirements and the dynamic changes of brain nerve activities.

[0005] At the same time, the existing brain nerve electrical signal acquisition technologies have deficiencies in accuracy and are difficult to capture weak nerve electrical signals or light signals. Brain nerve electrical signals have the characteristics of high frequency and high dimension, and existing signal processing methods often cannot perform real-time processing and analysis on the large amount of nerve electrical signals collected. This makes it impossible to obtain useful information in a timely manner when studying real-time brain nerve activities, and a large amount of time is required for offline processing, reducing the research efficiency. Most of the current signal analysis methods are based on traditional statistical and signal processing theories, and it is difficult to fully exploit the rich information contained in brain nerve electrical signals. For example, for complex neural oscillation signals and neural coding mechanisms, existing analysis methods may not be able to accurately analyze their internal neural activity patterns and functional significance.

[0006] In summary, there are many deficiencies in the existing technologies in optogenetic technology for the brain and signal processing. There is an urgent need for a two-dimensional on-chip brain light stimulation and signal processing technology to improve the accuracy and flexibility of optogenetic technology for the brain, enhance the ability of nerve electrical signal acquisition and processing, and provide more advanced technical support for brain neuroscience research and the treatment of nervous system diseases. Summary of the Invention

[0007] In order to solve the deficiencies of the above technical solutions, the purpose of the present invention is to provide a on-chip brain light stimulation system.

[0008] Another object of the present invention is to provide the signal processing method for on-chip brain optogenetic stimulation described above.

[0009] The object of the present invention is achieved by the following technical solutions.

[0010] An on-chip brain optogenetic stimulation system includes a light source module, a collimation module, a micro-projection module, a microscope module, a micro-projection driving module, an image acquisition module, an electrophysiological recording module, a stimulation and feedback recording module, and a data processing module.

[0011] The light emitted by the light source module is incident on the collimation module. The collimation module ensures that the light forms a parallel light beam and is incident on the micro-projection module. The micro-projection module controls through the micro-projection driving module to convert the parallel light beam into patterned light. The microscope module is used to project the patterned light onto the on-chip brain infected with adeno-associated virus to perform patterned light stimulation on the on-chip brain and generate a response image. The image acquisition module real-time collects the data of the patterned light stimulation. The electrophysiological recording module records the neural electrical signal data of the on-chip brain module. The stimulation and feedback recording module synchronizes the data of the patterned light stimulation and the neural electrical signal data in real time, feeds back to the micro-projection driving module to adjust the data of the patterned light stimulation, and performs data processing and analysis through the data processing module.

[0012] In the above technical solution, the stimulation and feedback recording module controls a data acquisition card (abbreviation: DAQ card) through LabVIEW software to output two TTL signals simultaneously. One TTL signal is transmitted to an electronic shutter, and the opening and closing of the electronic shutter are used to realize the switching of the micro-projection driving module. The other TTL signal is transmitted to the stimulation and feedback recording module.

[0013] In the above technical solution, the light source module includes a controller and a light source which are electrically connected. The controller is used to adjust the switching state and output intensity of the light source.

[0014] In the above technical solution, the collimation module includes an adjusting device and a collimation lens installed on the adjusting device. The adjusting device is used to adjust the position of the collimation lens to adapt to the incidence of the incident light beam, facilitating the refraction of the incident light beam. The collimation lens includes two plano-convex lenses, which are used to adjust the incident light beam into a parallel light beam to achieve the collimation of the light beam.

[0015] In the above technical solution, the micro-projection module includes a Digital Mirror Device (DMD) and a lens. The digital mirror device includes a plurality of micro-mirror units, and each micro-mirror unit is independently deflected by an electrical signal to form patterned light for stimulating the on-chip brain. The collimated parallel light beam passes through the plurality of micro-mirror units to generate patterned light, and then the patterned light is focused by the lens.

[0016] In the above technical solution, the micro-projection driving module includes a control circuit and a signal processor for controlling the independent deflection of the micro-mirror units to generate patterned light.

[0017] In the above technical solution, the microscope module includes a microscope optical system and optical elements. The microscope optical system is a microscope with a side-connected light source for projecting the patterned light onto the on-chip brain.

[0018] In the above technical solution, the electrophysiological recording module includes an MEA multi-electrode array, a signal amplifier, and a data acquisition system for collecting and recording the neural electrical signals of the on-chip brain. The electrode chip of the MEA multi-electrode array is coated and connected to the on-chip brain to form a biochip for high-throughput electrophysiological recording to obtain neural electrical signals.

[0019] In the above technical solution, the gene of the adeno-associated virus is pAAV-hSyn-ChrimsonR-EGFP-WPRE, and the adeno-associated virus carries a protein that specifically expresses a light-sensitive channel.

[0020] Another aspect of the present invention further includes a signal processing method for on-chip brain light stimulation, including the following steps:

[0021] Step 1: Use the on-chip brain light stimulation system to apply multiple rounds of light stimulation with a set pattern to the on-chip brain, and perform multiple rounds of full-dark processing on the on-chip brain, and collect neural electrical signals;

[0022] Step 2: Perform phase space reconstruction on the neural electrical signals by the phase space reconstruction method to obtain the signal characteristics in the high-dimensional space, so as to obtain the dynamic state of the on-chip brain neuron cells under the condition of light stimulation through the phase space. Specifically, it includes the following steps:

[0023] Step 2.1: Determine the optimal delay time for phase space reconstruction: Use the C-C algorithm based on Logistics to calculate the correlation function in the embedded multi-dimensional phase space and the correlation function in the one-dimensional phase space respectively, then calculate the difference between multiple groups of correlation functions in the embedded multi-dimensional phase space and the correlation function in the one-dimensional phase space, and calculate the average value of the differences. Judge the optimal delay time τ according to the average value of the differences. Among them, the correlation function is:

[0024]

[0025] Among them, m is the dimension of the embedded phase space, N is the length of the neural electrical signal, r is the threshold. The standard deviations of multiple groups of neural electrical signals are taken and averaged. t is the length of the neural electrical signal divided into t segments. After the length N of the neural electrical signal is divided into t segments, one of the segments is embedded into the m-dimensional phase space, and the phase point in the phase space obtained is X, X i and X j are different phase points in the phase space, M is the number of phase points in the phase space, θ(·): if the content in the parentheses is negative, the correlation function is 0, otherwise it is 1;

[0026] For each segmented segment s, there corresponds a correlation function C, and the average value of the defined difference is:

[0027]

[0028] Select the first minimum point of the S value, and the value of its corresponding abscissa is the optimal delay time τ;

[0029] Step 2.2, determine the optimal dimension of the phase space reconstruction:

[0030] Define the Logistics autocorrelation function: Among them, x represents the discrete time series, Z represents the total length of the time series x, k represents the delay of the time series x. The autocorrelation function values with delay times from 1 to 200 are calculated respectively, the relationship diagram between the delay time and the autocorrelation function value is plotted, and the first minimum point in this diagram is taken, then its abscissa is the optimal dimension of the phase space reconstruction;

[0031] Step 2.3, construct the phase space by using the optimal delay time τ in Step 2.1 and the optimal dimension in Step 2.2, and analyze this phase space by downsampling the neural electrical signal collected in Step 1 to 1000 Hz. 0.25 times the standard deviation of the neural electrical signal is used as the threshold to draw the recurrence plot. In the constructed phase space, if the Euclidean distance between two points in the recurrence plot is less than the threshold, a point is marked at the position of the labels of these two points in the recurrence plot. If it is greater than the threshold, it is not marked on the plot. Assume that the function value corresponding to the marked point is 1, and the function value without marking is 0. The function expression used for the function value is:

[0032]

[0033] Among them, is an element of the recurrence matrix, ‖X i -X j ‖ is the phase point X i and the phase point X jThe distance between them, where θ is the Heaviside step function;

[0034] Step 2.4: After performing Fourier transform on the neural electrical signal, find the modulus length (i.e., power) of each component, and identify the frequency corresponding to the position with the maximum power. The reciprocal of this frequency is the relative period T. In the phase space, find the nearest neighbor phase point X i within a non-one period T of X j , that is:

[0035] d i (0) = min‖X i - X j ‖ where |i - j| > T

[0036] For the phase point X i after evolving to the nth point, the phase point X i+n , and at the same time the phase point X j after evolving to the nth point, the phase point X j+n , the Euclidean distance between the phase point X i+n and the phase point X j+n is:

[0037] d i (n) = ‖X i+n - X j+n ‖ where n ≤ min{M - i, M - j}

[0038] where M is the total number of phase points in the phase space. Since there is the following relationship between the distance from the initial phase point after evolving to the nth phase point in the phase space: Taking the logarithm on both sides can obtain a linear relationship. Then, evolve all the phase points in the phase space, take the logarithm of the value of the distance d i (n) for each phase point n, and use the formula <d i (n)> i = <d i (0)> i + λ1(nΔt), calculate the average value, and fit all the average values into a straight line. The slope of this straight line is the Lyapunov exponent. Use the Lyapunov exponent to explore the initial value sensitivity of on-chip brain neurons under light treatment and full-dark treatment;

[0039] Step 2.5: Perform principal component analysis on the neural electrical signals collected on one electrode for one round of light stimulation and one round of full-dark treatment, perform orthogonal transformation on the high-dimensional phase points in the phase space, arrange the features in the high-dimensional phase points in descending order of variance, and draw a trajectory diagram using the first two principal components with the largest variance, so as to observe the changes in the on-chip brain dynamics trajectories under two different states.

[0040] The advantages and beneficial effects of the present invention are:

[0041] 1. The system of the present invention uses the light beam emitted by the light source module to be collimated by the collimation module, transmits the incident light beam to the digital micromirror device through the collimation system, generates a patterned optogenetic stimulation image by controlling the deflection of multiple micromirror units, projects the image onto the on-chip brain through the microscope optical path for optogenetic stimulation, uses the image acquisition module to achieve real-time observation, and records the two-dimensional on-chip brain nerve electrical signals through the electrodes of the multi-electrode array, and performs time-domain, frequency-domain and dynamic analysis on the signals recorded in real time. Realize the real-time synchronization of optogenetic stimulation and electrical signal recording for the two-dimensional on-chip brain.

[0042] 2. The Lyapunov exponent of the present invention is used to measure the input of an initial value with a small perturbation into the dynamic system. This small perturbation evolves in the system and brings changes to the output after a period of time. The growth rate of the change of this small perturbation in the system is called the Lyapunov exponent, which reflects the sensitivity of the system to the initial conditions. In a chaotic system, a small initial perturbation will lead to a huge difference in the system evolution, and this sensitivity is also one of the important characteristics of the chaotic phenomenon.

[0043] 3. The recurrence plot of the present invention is a powerful means to represent the dynamic characteristics of signals. For non-linear signals, they often exhibit chaotic characteristics, which are shown as black regions appearing in blocks on the recurrence plot. For a completely chaotic signal, a system without self-organization ability, its signal is chaotic, and the shape shown on the recurrence plot is also scattered points without a distribution pattern. Brief Description of the Drawings

[0044] Figure 1 It is a flow chart for preparing the two-dimensional light-stimulated on-chip brain in Example 1.

[0045] Figure 2 It is a fluorescence gene expression map of the optogenetic neuron vector observed by a confocal microscope 11 days after transfection of the adeno-associated virus in Example 1. Among them, A is the neuron map in the normal light mode, and B is the neuron map in the green fluorescence mode.

[0046] Figure 3 It is a framework diagram of the on-chip brain light stimulation system in Example 2.

[0047] Figure 4 It is a schematic structural diagram of the on-chip brain light stimulation system in Example 2.

[0048] Figure 5 It is a light stimulation optical path diagram in Example 2.

[0049] Figure 6 It is a schematic diagram of the light stimulation process in Example 3. Among them, A is Stimulation Scheme 1, and B is Stimulation Scheme 2.

[0050] Figure 7Line graph showing the relationship between the classification accuracy of the support vector machine and time.

[0051] Figure 8 Spike distribution raster plot for Example 3, where A is the raster of network bursts and B is the raster of evoked potentials with light stimulation.

[0052] Figure 9 Neuroelectrophysiological signal waveform diagram for Example 3, where A is the spontaneous neuroelectrical signal and B is the neuroelectrical signal evoked by light stimulation.

[0053] Figure 10 Principal component analysis diagram for 60 channels.

[0054] Figure 11 Lyapunov number - number of stimuli line graph.

[0055] Among them, 110: light source module, 111: controller, 112: light source, 120: collimation module, 121: plano-convex lens, 122: plano-convex lens, 130: micro-projection module, 131: digital micromirror device, 132: lens, 140: micro-projection drive module, 150: microscope module, 160: image acquisition module, 161: camera, 170: on-chip brain, 180: electrophysiological recording module, 181: MEA electrode, 190: stimulation and feedback recording module, 200: data processing module. Detailed implementation manners

[0056] The technical solution of the present invention will be further described below in conjunction with specific embodiments.

[0057] Example 1

[0058] As Figure 1 shown, a method for preparing a two-dimensional on-chip brain includes the following steps:

[0059] Step 1, obtain a suspension of cortical neurons (cortical cells) of a mouse. After digestion, add 4 ml of planting medium (DMEM / F12 + 10% FBS + 1% P.S.) to terminate digestion. Filter the cell suspension using a 200-mesh filter to filter out impurities in the suspension, obtain a single-cell suspension of neurons, then centrifuge at 1000 rpm for 5 minutes, discard the supernatant, and resuspend with 1 ml of planting medium to obtain a resuspended nerve cell suspension.

[0060] Step 2, after counting the resuspended nerve cell suspension, dilute it with the planting medium to a concentration of 1.5*10 6cells / ml. Seed 200 μL of cell suspension on the MEA multi-electrode array, and place the well plate and the multi-electrode array in a cell culture incubator at 37 °C, which is recorded as day 0 of in vitro neuron culture. After 4 hours, perform a full medium change and replace it with neuron maintenance medium (Neurobasal + 2% B27plus + 1% P.S. + 0.5 mM Glutamax). Thereafter, every 3 days, perform a half medium change of the neuron maintenance medium.

[0061] Step 3, after 10 days of in vitro neuron culture, infect the neurons with an adeno-associated virus carrying a light channel protein to enable the neurons to express the light channel protein, so that the finally generated on-chip brain in two dimensions exhibits electrophysiological characteristics in response to light stimulation (as Figure 2 shown, the neurons appear green in the figure, representing the fluorescent protein on their expression vector, and approximately 60 - 70% of the neurons express this gene). Among them, the adeno-associated virus is H15951 from OBiO, packaged as AAV2 / 1 (AAV2 genome, AAV1 capsid protein), and its gene is pAAV-hSyn-ChrimsonR-EGFP-WPRE: pAAV represents plasmid AAV, that is, the adeno-associated virus plasmid vector; hSyn represents human synapsin, that is, the human synaptic-specific neuron marker gene, which is a gene encoding a synaptic vesicle protein and is mainly expressed in neurons and serves as a promoter in the plasmid to drive gene expression; ChrimsonR represents the gene encoding the light channel protein in this embodiment. Under the irradiation of blue light (470 nm), the channel opens, and sodium ions or calcium ions in the culture medium enter the cell; EGFP represents Enhanced Green Fluorescent Protein, which will produce green fluorescence under the excitation of green light; WPRE is an enhancer derived from the woodchuck hepatitis virus, which is used to improve the transcription efficiency and the stability of mRNA. In this embodiment, since it is located downstream of the light channel protein on the plasmid, by observing the expression of the green fluorescent protein, the expression of the light channel protein can be estimated, and thus the transfection of the adeno-associated virus can be estimated. The specific steps are as follows:

[0062] Step 3.1, dilute the adeno-associated virus with the maintenance medium, completely aspirate the original medium of the neurons seeded on the MEA multi-electrode array, and change it to the maintenance medium containing the virus. Infect the neurons with the adeno-associated virus at a ratio of 1:10 5 After 24 hours, completely aspirate the maintenance medium containing the virus and change it to the medium without the virus.

[0063] Step 3.2: The neurons are cultured in vitro for another 7 to 10 days until the neurons mature and are interconnected to form a network. When observing the electrical signals of the neurons using a multi-electrode array (MEA), it is found that there is a phenomenon of network bursts in multiple channels simultaneously, forming a mature two-dimensional on-chip brain infected with adeno-associated virus, which contains the light channel gene (ChrimsonR) and the enhanced green fluorescent gene (EGFP). When observing with a confocal microscope, the green fluorescence is obvious and the light channel protein is fully expressed, then this two-dimensional on-chip brain can be used for light stimulation experiments.

[0064] Example 2

[0065] As Figures 3 - 5 shown, this example provides an on-chip brain light stimulation system, which is used to apply light stimulation to the two-dimensional on-chip brain prepared in Example 1 and to collect and analyze neural electrical signals. The on-chip brain light stimulation system includes a light source module 110, a collimation module 120, a micro-projection module 130, a micro-projection driving module 140, a microscope module 150, an image acquisition module 160, an electrophysiological recording module 180, a stimulation and feedback recording module 190, and a data processing module 200;

[0066] The light emitted by the light source module 110 is incident on the collimation module. The collimation module 120 ensures that the light maintains a consistent light intensity and beam size during propagation, and finally forms a parallel beam of light incident on the micro-projection module 130. The micro-projection module 130 controls the parallel beam of light to be converted into patterned light through the micro-projection driving module 140. The microscope module 150 is used to project the patterned light onto the two-dimensional on-chip brain infected with adeno-associated virus obtained in Example 1, apply patterned light stimulation to the on-chip brain 170, generate a response image, the image acquisition module 160 real-time collects the data of the patterned light stimulation, the electrophysiological recording module 180 records the neural electrical signal data of the on-chip brain module, and the stimulation and feedback recording module 190 synchronizes the data of the patterned light stimulation and the neural electrical signal data in real time, feeds back to the micro-projection driving module to adjust the data of the patterned light stimulation and performs data processing and analysis through the data processing module 200;

[0067] The stimulation and feedback recording module 190 controls a data acquisition card (abbreviated as DAQ card) through LabVIEW software to output two TTL signals simultaneously. One TTL signal is transmitted to the electronic shutter, and the opening and closing of the electronic shutter are used to realize the on-off of the micro-projection driving module; the other TTL signal is transmitted to the stimulation and feedback recording module;

[0068] The entire program mainly includes two processes: event operation and process operation, which are used to handle the operations of users on the operation interface and design the program process respectively, and realize the information transmission between these two parts through queue operation. Create a DAQ channel - start a DAQ task - write data to the DAQ - clear the DAQ task, so as to output the generated neural electrical signals through a data acquisition card.

[0069] Furthermore, the light source module 110 includes a controller 111 and a light source 112 that are electrically connected. The controller 111 is used to adjust the on / off state and output intensity of the light source 112. The central wavelength of the light source is 470 nm to match the photosensitive protein ChR2.

[0070] The collimation module 120 includes an adjustment device and a collimation lens installed on the adjustment device. The adjustment device is used to adjust the position of the collimation lens to adapt to the incidence of the incident light beam, facilitating the refraction of the incident light beam. The collimation lens includes two plano-convex lenses 121 / 122; the light beam emitted by the light source has a certain divergence angle. After passing through the collimation lens, according to the law of refraction, after the incident light beam is refracted by the two plano-convex lenses, the divergence angle decreases, and the incident light beam is adjusted into a parallel light beam, realizing the collimation of the light beam.

[0071] The micro-projection module 130 includes a digital micromirror device 131 (Digital Mirror Device, DMD) and a lens 132. The digital micromirror device contains a plurality of micromirror units. Each micromirror unit deflects independently through an electrical signal (specifically, the micromirror unit deflects ±12° along the diagonal direction, and the reflected light forms an angle of 24° with the vertical plane in the horizontal axis 45° direction) to form patterned light to stimulate the on-chip brain; the collimated parallel light beam passes through a plurality of micromirror units to generate patterned light, and then the patterned light is focused through the lens. Specifically, by customizing a black-and-white binary picture and loading it into the DMD through a broadband connection method, the patterning of the light stimulation is realized. By clicking the icon in real time on the Matlab platform and returning the time of the icon click, the synchronization of the light stimulation and the neural electrical signal is realized; the micro-projection driving module 140 includes a control circuit and a signal processor, which are used to control the independent deflection of the micromirror units to generate patterned light.

[0072] The microscope module 150 includes a microscope optical system and optical elements. The microscope optical system is a microscope with a side-connected light source, which is used to project the patterned light onto the on-chip brain.

[0073] The electrophysiological recording module 180 includes an MEA multi-electrode array 181, a signal amplifier, and a data acquisition system, which are used to collect and record the neural electrical signals of the on-chip brain. The chip of the MEA electrode is connected to the on-chip brain using the method described in Example 1 to form a biochip. (The on-chip brain is transfected with a light channel protein to make it specifically responsive to patterned light stimulation), and high-throughput electrophysiological recording is performed to obtain rich neural activity data (neural electrical signals).

[0074] Specifically, the MEA electrode is a multi-electrode array (Multi Electrode Arrays, MEAs) from the MCS company, which is used to detect and record the extracellular spike potentials at multiple sites of the in vitro biological neural network. This MEAs has a total of 59 TiN / SiN electrodes, arranged in an 8*8 array, excluding 4 top corners and the reference electrode. There are a total of 59 electrodes that can detect the extracellular potential of the neural network, corresponding to 60 channels.

[0075] Example 3

[0076] As Figure 6 shown, this embodiment provides a method for optically stimulating the two-dimensional on-chip brain infected with adeno-associated virus in Example 1 using the on-chip brain optical stimulation system in Example 2, including two stimulation schemes.

[0077] As Figure 6 shown in A of

[0078] Step 1, first use the MEA multi-electrode array to record the spontaneous electrical signals of the on-chip brain for 7 - 8 seconds to observe whether the electrophysiological state of the on-chip brain is normal.

[0079] Step 2, within 200 ms, randomly perform a full-dark treatment on the on-chip brain (without shining light on the three-dimensional on-chip brain), or use a digital micromirror device to optically stimulate the entire on-chip brain, and then perform a full-dark treatment for 5 s. This is one stimulation cycle, and the stimulation cycle is repeated 20 times. During the entire cyclic stimulation process, the full-bright treatment for optical stimulation appears 10 times, and the full-dark treatment appears 10 times.

[0080] The purpose of Stimulation Scheme 1 is to verify whether the adeno-associated virus has successfully transfected the neuron light channel protein in the two-dimensional on-chip brain, and whether the on-chip brain optical stimulation system in Example 2 can successfully open the light channel protein, thereby causing changes in the neuron membrane potential. It is found that the form of the neuron membrane potential change caused by this on-chip brain optical stimulation system is somewhat different from the neuron spontaneous electricity in form.

[0081] As Figure 6 shown in B of

[0082] Step 1: Record the spontaneous electrical signals of the on-chip brain for 7 - 8 seconds to observe whether the electrophysiological state of the on-chip brain is normal.

[0083] Step 2: First, within 200 ms, control the digital micromirror device (DMD) to randomly apply stimuli of full brightness, full darkness, and the letter 'A' to the on-chip brain. Then, perform full darkness treatment for 5 s. This is one stimulation cycle, and repeat the stimulation cycle 30 times, ensuring that during the entire cycle of stimulation, full brightness treatment occurs 10 times, full darkness treatment occurs 10 times, and the letter 'A' stimulation treatment occurs 10 times.

[0084] This stimulation scheme aims to verify whether the neural network will produce different responses to stimuli of full darkness, full brightness, and patterned stimuli.

[0085] Example 4

[0086] This example provides a method for processing the signals collected by the multi - electrode array after stimulating the on-chip brain with the optical stimulation scheme 1 described in Example 3, including the following steps:

[0087] Processing method 1:

[0088] Step 1: The multi - electrode array has 60 channels, and each channel has 60 columns of signals. Taking the 1 s of signals before 200 ms and the 4 s of signals after 200 ms centered around 200 ms for each column of signals to form a signal slice.

[0089] Step 2: In the optical stimulation scheme 1, full brightness treatment occurs 10 times and full darkness treatment occurs 10 times. During the optical stimulation process, the multi - electrode array can collect 20 groups of signal slices. Select the first 14 signal slices as the training data set and the last 6 signal slices as the test data set to train and test the support vector machine, and then use the support vector machine to classify the signal slices to verify the electrophysiological state of the two - dimensional on-chip brain in the two different states of full brightness and full darkness during optical stimulation.

[0090] As Figure 7As shown, in step 3, a signal slice (5.2 s) is evenly divided with a 50-ms sliding time window, and the number of spikes of the neural electrical signals within all the sliding time windows is counted. The minimum time point in each sliding time window is the time corresponding to the number of spikes. Finally, a feature signal is obtained. The feature signal, the corresponding time, and the electrode position corresponding to obtaining this feature signal are used as the input of a Support Vector Machine (SVM). Each sliding time window with light stimulation is labeled as 1, and without light stimulation is labeled as 0, forming a time-related binary labeling sequence as the output of the SVM. The training set and test set in step 2 are used to train and test the SVM, and the classification accuracy of the SVM is used to evaluate the firing patterns of two-dimensional on-chip brain neurons in two states of full-light processing or full-dark processing. Among them, the classification accuracy is the number of times correctly classified by the SVM divided by all the data used for testing. According to the results of the classification accuracy rate, between 100 - 200 ms after the light stimulation occurs, the SVM classification starts to have specificity (classification accuracy greater than 50%), which indicates that there is a delay of more than 100 ms between the start of light stimulation and the start of neurons responding to the stimulation; after about 250 ms, the maximum value of the classification accuracy is reached; after about 500 ms, the specificity of this classification disappears (classification accuracy less than or equal to 50%).

[0091] As can be seen Figure 7 from the figure, as described in step 2, the first 14 times are for training, and the last 6 times are for testing. The total number of spikes occurring within is counted with a 50-ms time window, and the classification accuracy of the SVM for the last 6 times is calculated. The blue curve is the maximum value considering classification with time and space factors (1.25, 0.8830). The time of the gray part is the time period when full-light stimulation occurs, from 1.15 s to 1.35 s, lasting for 200 ms.

[0092] Processing method two:

[0093] As Figure 8As shown in the figure, a fourth-order Butterworth bandpass filter is used to extract the high-frequency part of the signal with a frequency range of 300 - 3000 Hz in the signal slice. The variance of this part of the signal is calculated with time as the variable (that is, the average value of the signal is calculated according to time, and then for all signal values at different times, the difference between each signal value and this average value is squared, and then the average value of all squared differences is calculated to obtain the variance). When the peak value of the signal is greater than -5 times the variance and there is no second peak value within 2 ms (the absolute refractory period of cortical neurons is about 2 ms), it indicates that an action potential has been generated in the two-dimensional on-chip brain neurons on this electrode. The spike value at this moment is 1, otherwise the spike value at this moment is 0. The data series obtained through this process belongs to a binary signal of 0 - 1, and this signal is called a spike signal. A program is written using Matlab to visualize the spike signal. A rectangular coordinate system is established, where the horizontal axis of the rectangular coordinate system represents time and the vertical axis represents channels. When the spike value at a certain time is 1, a point is plotted at the corresponding position, thus plotting the spike signals of 60 channels into a neural electrical signal raster plot.

[0094] By optically stimulating the two-dimensional on-chip brain with the on-chip brain optical stimulation system in Example 2, the MEA multi-electrode array collects neural electrical signals, and a spike distribution raster plot of 60 channels is obtained. As Figure 8 shown, if there is a spike in this channel at this time, a point is marked on the graph. Figure 7 In Figure B of , the red vertical lines indicate the start and end times of the optical stimulation. It can be seen that the network burst is different from the spike firing pattern of the evoked potential of the optical stimulation.

[0095] Processing method three:

[0096] As Figure 9 shown, a 400th-order Fir bandpass filter is used to extract the low-frequency part of the electrical signal with a frequency range of 1 - 300 Hz in the signal slice. For this part of the signal, a 20000th-order Fir bandpass filter is used to extract the delta wave of 1 - 4 Hz; a 10000th-order Fir bandpass filter is used to extract the theta wave of 4 - 11 Hz; a 10000th-order Fir bandpass filter is used to extract the beta wave of 11 - 30 Hz; a 5000th-order Fir bandpass filter is used to extract the gamma wave of 30 - 55 Hz, and at the same time, the waveform diagram of the spontaneous electrical signal of the two-dimensional on-chip brain is plotted.

[0097] As Figure 9 shown, from top to bottom are the original signal, delta wave, theta wave, beta wave, and gamma wave.

[0098] As Figures 10 - 11As shown, Method 4, which can indirectly reveal the plasticity of short-term memory neurons, includes the following steps:

[0099] Step 1, collect the neural electrical signals obtained by Stimulation Scheme 1;

[0100] Step 2, perform phase space reconstruction on the neural electrical signals through the phase space reconstruction method to obtain the signal features in the high-dimensional space, so as to obtain the dynamic state of on-chip brain neurons under the condition of light stimulation through the phase space. Specifically, it includes the following steps:

[0101] Step 2.1, determine the optimal delay time for phase space reconstruction: Use the C-C algorithm based on Logistics to calculate the correlation function of the embedded multi-dimensional phase space and the correlation function of the one-dimensional phase space respectively, then calculate the difference between multiple groups of correlation functions of the embedded multi-dimensional phase space and the one-dimensional phase space, and calculate the average value of the differences. Determine the optimal delay time τ according to the average value of the differences. Among them, the correlation function is:

[0102]

[0103] where m is the dimension of the embedded phase space, and in this embodiment, the average value in four cases is obtained by taking values from 2 to 5. N is the length of the neural electrical signal, r is the threshold. In this embodiment, r takes 0.5 times, 1 time, 1.5 times, and 2 times the standard deviation of the signal respectively, and the average value is obtained. t is the neural electrical signal length divided into t segments. After the neural electrical signal length N is divided into t segments, one of the segments is embedded into the m-dimensional phase space, and the phase point in the phase space is X, X i and X j are different phase points in the phase space, M is the number of points in the phase space, θ(·): If the content in the parentheses is negative, the correlation function is 0, otherwise it is 1;

[0104] For each segmentation segment s, there corresponds a correlation function C. The C-C algorithm compares the difference when the embedding is m-dimensional and when the embedding is 1-dimensional, and defines the average value of the differences:

[0105]

[0106] Select the first minimum point of the S value, and the corresponding abscissa value is the optimal delay time τ.

[0107] Step 2.2, determine the optimal dimension of phase space reconstruction (using the method of the autocorrelation function based on Logistics, the advantage of this method is good robustness and it is applicable to small data volumes):

[0108] Define the Logistics autocorrelation function: Among them, \(x\) represents a discrete time series, \(Z\) represents the total length of the time series \(x\), and \(k\) represents the delay of the time series \(x\). Calculate the autocorrelation function values with delays ranging from 1 to 200 respectively, plot the relationship between the delay time and the autocorrelation function values, and take the first minimum point in this graph. Then its abscissa is the optimal dimension for phase space reconstruction.

[0109] Step 2.3: Construct a phase space using the optimal delay time \(\tau\) in Step 2.1 and the optimal dimension in Step 2.2. Analyze this phase space by downsampling the neural electrical signals collected in Step 1 to 1000 Hz. Use 0.25 times the standard deviation of the neural electrical signals as the threshold \(r\) to draw a recurrence plot. In the constructed phase space, if the Euclidean distance between two points in the recurrence plot is less than the threshold, mark a point at the position of the labels of these two points in the recurrence plot; if it is greater than the threshold, do not mark on the graph. Assume that the function value corresponding to the marked point is 1, and the function value without marking is 0. Among them, the function expression used for the function value (for recurrence plot analysis) is:

[0110]

[0111] Among them, is an element of the recurrence matrix, \(\|X\) i -X\) j \| is the distance between the phase point \(X\) i and the phase point \(X\) j , and \(\theta\) is the Heaviside step function.

[0112] Step 2.4: Calculate the Lyapunov exponent using the small data method of Lyapunov. Calculate the Lyapunov exponent using the small data method. First, embed the neural electrical signals into the phase space in Step 2.3. In the phase space, find the nearest neighbor points of each phase point within one period. This period is calculated using the power spectrum method, which specifically includes the following steps: After performing a Fourier transform on the neural electrical signals, calculate the modulus length of each component, which is the power, and find the frequency corresponding to the position with the maximum power. The reciprocal of this frequency is the relative period \(T\). In the phase space, find the nearest neighbor points \(X\) i of each phase point \(X\) j within one period \(T\), that is:

[0113] d i (0)=\min\{\|X\) i -X\) j \|\mid i - j\gt T\)

[0114] For the phase point \(X\) i evolved to the phase point \(X\) i+n after the \(n\)th point, and at the same time the phase point \(X\) j evolved to the phase point \(X\) j+n after the \(n\)th point, the phase point \(X\)i+n The Euclidean distance to the phase point X j+n is:

[0115] d i (n) = ‖X i+n - X j+n ‖ where n ≤ min{M - i, M - j}

[0116] where M is the total number of phase points in the phase space. Since there is the following relationship between the distance of the nth phase point evolved in the phase space and the distance from the initial phase point: Taking the logarithm on both sides can obtain a linear relationship. Then, all phase points in the phase space are evolved. For the distance d i (n) of each phase point n, take the logarithm of its value, and use the formula <d i (n)> i = <d i (0)> i + λ1(nΔt), calculate the average value of all phase points, fit all the average values into a straight line. The slope of this straight line is the Lyapunov exponent. Use the said Lyapunov exponent to explore the initial value sensitivity of on-chip brain neurons under full-brightness processing and full-darkness processing.

[0117] Step 2.5, Component analysis of the phase space: According to the phase space reconstruction, high-dimensional phase points in the phase space can be obtained. If the components of each phase point are regarded as one feature and the neural electrical signal is embedded into M dimensions, each phase point will have M features. Therefore, principal component analysis can be performed on these phase points. The method of principal component analysis is to perform an orthogonal transformation on the high-dimensional phase points, and then arrange their features in descending order of variance to better unfold the trajectory of the phase points in the phase space. The specific steps are as follows: Perform principal component analysis on the neural electrical signals collected on one electrode for one round of light stimulation and one round of full-darkness processing, perform an orthogonal transformation on the high-dimensional phase points in the phase space, arrange the features in the high-dimensional phase points in descending order of variance, and draw a trajectory diagram using the first two principal components with the largest variances, so as to observe the changes in the on-chip brain dynamics trajectories under two different states.

[0118] As can be seen from Figure 10 the red line shows that the 60 channels of the on-chip brain are in a chaotic motion state in the phase space without light stimulation, and the blue line shows the phase space trajectory of the on-chip brain under light stimulation, where the trajectory deviates from the original resting state and then returns to the original state after one week.

[0119] As can be seen from Figure 11It can be seen that the blue broken line represents the Lyapunov number without light stimulation (full dark treatment). It can be observed that its response to the initial value perturbation remains almost unchanged, indicating that the network is in a steady state. The red broken line represents the on-chip brain initial value response under global light stimulation. The sensitivity of the on-chip brain to the initial value fluctuates, and generally increases with the increase in the number of stimulations, indicating that the on-chip brain is in a perturbed state or a metastable state at this time.

[0120] It should be noted that the processing method of this embodiment is also applicable to Stimulation Scheme 2 of Embodiment 3.

[0121] The above provides an exemplary description of the present invention. It should be noted that without departing from the core of the present invention, any simple deformation, modification, or equivalent replacement that can be made by those skilled in the art without creative efforts falls within the protection scope of the present invention.

Claims

1. An on-chip brain optical stimulation system, characterized in that, It includes a light source module, a collimation module, a micro-projection module, a microscope module, a micro-projection driving module, an image acquisition module, an electrophysiological recording module, a stimulation and feedback recording module, and a data processing module; The light emitted by the light source module is incident on the collimation module. The collimation module ensures that the light forms a parallel light beam and is incident on the micro-projection module. The micro-projection module, controlled by the micro-projection driving module, converts the parallel light beam into patterned light. The microscope module is used to project the patterned light onto the on-chip brain infected with adeno-associated virus to perform patterned light stimulation on the on-chip brain. The image acquisition module real-time collects the data of the patterned light stimulation. The electrophysiological recording module records the neural electrical signal data of the on-chip brain module. The stimulation and feedback recording module synchronizes the patterned light stimulation data and the neural electrical signal data in real time, feeds them back to the micro-projection driving module to adjust the patterned light stimulation data, and performs data processing and analysis through the data processing module.

2. The on-chip brain optical stimulation system according to claim 1, wherein The stimulation and feedback recording module controls the data acquisition card through LabVIEW software to output two TTL signals simultaneously. One TTL signal is transmitted to the electronic shutter, and the on / off of the electronic shutter is used to realize the on / off of the micro-projection driving module; the other TTL signal is transmitted to the stimulation and feedback recording module.

3. The on-chip brain optogenetic stimulation system according to claim 1, characterized in that, The light source module includes a controller and a light source connected electrically. The controller is used to adjust the on / off state and output intensity of the light source.

4. The on-chip brain optical stimulation system according to claim 1, characterized in that, The collimation module includes an adjustment device and a collimation lens installed on the adjustment device. The adjustment device is used to adjust the position of the collimation lens. The collimation lens includes two plano-convex lenses and is used to adjust the incident light beam into a parallel light beam to realize the collimation of the light beam.

5. The on-chip brain optogenetic stimulation system according to claim 1, characterized in that, The micro-projection module includes a digital micromirror device and a lens. The digital micromirror device contains multiple micromirror units. Each micromirror unit deflects independently through an electrical signal to form patterned light to stimulate the on-chip brain; the collimated parallel light beam passes through multiple micromirror units to generate patterned light, and then the patterned light is focused through the lens.

6. The on-chip brain optical stimulation system according to claim 1, wherein The micro-projection driving module includes a control circuit and a signal processor, and is used to control the independent deflection of the micromirror units to generate patterned light.

7. The on-chip brain optical stimulation system according to claim 1, characterized in that The microscope module includes a microscope optical system and optical elements. The microscope optical system is a microscope with a side-connected light source and is used to project the patterned light onto the on-chip brain.

8. The on-chip brain optical stimulation system according to claim 1, characterized in that, The electrophysiological recording module includes an MEA multi-electrode array, a signal amplifier, and a data acquisition system, and is used to collect and record the neural electrical signals of the on-chip brain. The electrode chip of the MEA multi-electrode array is coated and connected to the on-chip brain to form a biochip for high-throughput electrophysiological recording to obtain neural electrical signals.

9. The on-chip brain optical stimulation system according to claim 1, wherein The gene of the adeno-associated virus is pAAV-hSyn-ChrimsonR-EGFP-WPRE, and the adeno-associated virus carries a protein that specifically expresses a light-sensitive channel.

10. A method for processing on-chip brain light stimulation signals, comprising the following steps: Step 1, using the on-chip brain light stimulation system according to any one of claims 1 to 9 to apply multiple rounds of light stimulation with a set pattern to the on-chip brain, and performing multiple rounds of full-dark treatment on the on-chip brain, and collecting neural electrical signals; Step 2: Reconstruct the phase space of the neuroelectrical signals through the phase space reconstruction method to obtain the signal features in the high-dimensional space, so as to obtain the dynamic state of the on-chip brain neurons under the condition of light stimulation. The specific steps are as follows: Step 2.1: Determine the optimal delay time for phase space reconstruction. Calculate the correlation function of the multi-dimensional phase space and the correlation function of the one-dimensional phase space respectively using the C-C algorithm based on Logistics, then calculate the difference between the correlation functions of multiple groups of multi-dimensional phase spaces and one-dimensional phase spaces, and calculate the average value of the differences. Determine the optimal delay time τ according to the average value of the differences. The correlation function is: Among them, m is the dimension of the embedded phase space, N is the length of the neural electrical signal, r is the threshold, the standard deviations of multiple groups of neural electrical signals are taken and the average value is obtained, t is the length of the neural electrical signal divided into t segments. After the length N of the neural electrical signal is divided into t segments, one of the segments is embedded into the m-dimensional phase space, and the phase point in the phase space obtained is X, X i and X j are different phase points in the phase space, M is the number of phase points in the phase space, θ(·): if the value inside the parentheses is negative, the correlation function is 0, otherwise it is 1; For each segmented segment s, there corresponds a correlation function C. Define the average value of the differences: Select the first minimum point of the S value, and the abscissa value corresponding to it is the optimal delay time τ; Step 2.2: Determine the optimal dimension of phase space reconstruction; Define the Logistics autocorrelation function: Among them, \(x\) represents the discrete time series, \(Z\) represents the total length of the time series \(x\), \(k\) represents the delay of the time series \(x\). Calculate the autocorrelation function values with the delay time ranging from 1 to 200 respectively, plot the relationship diagram between the delay time and the autocorrelation function values, and take the first minimum point in this diagram. Then its abscissa is the optimal dimension of phase space reconstruction; Step 2.3: Construct a phase space using the optimal delay time τ in Step 2.1 and the optimal dimension in Step 2.

2. Analyze this phase space by downsampling the neuroelectrical signals collected in Step 1 to 1000 Hz. Use 0.25 times the standard deviation of the neuroelectrical signals as the threshold to draw a recurrence plot. In the constructed phase space, if the Euclidean distance between two points in the recurrence plot is less than the threshold, mark a point at the position of the labels of these two points in the recurrence plot; if it is greater than the threshold, do not mark on the plot. Assume that the function value corresponding to the marked point is 1, and the function value without marking is 0. The function expression used for the function value is: Among them, is an element of the recursive matrix, ‖X i -X j ‖ is the distance between the phase points X i and the phase point X j ; θ is the Heaviside step function. Step 2.4, after performing Fourier transform on the neural electrical signals, calculate the modulus length (i.e., power) of each component, find the frequency corresponding to the position with the maximum power, and the reciprocal of this frequency is the relative period T. Then, find the nearest neighbor phase point X i within a non-one period T of j , that is: d i (0) = min ‖X i -X j ‖ |i - j| > T For phase point X i Phase point X after evolving to the nth point i+n , and at the same time phase point X j , phase point X after evolving to the nth point j+n , phase point X i+n and phase point X j+n The Euclidean distance between them is: d i (n) = ‖X i+n -X j+n ‖n ≤ min{M - i, M - j} Among them, M is the total number of phase points in the phase space. Since the distance between the n-th phase point evolved in the phase space and the initial phase point has the following relationship: Taking the logarithm on both sides can obtain a linear relationship. Then, all phase points in the phase space are evolved, and the logarithm of the distance d i (n) is taken. Using the formula <d i (n)> i = <d i (0)> i + λ1(nΔt), the average value is obtained, and all average values are fitted into a straight line. The slope of this straight line is the Lyapunov exponent. The Lyapunov exponent is used to explore the initial value sensitivity of on-chip brain neurons under bright treatment and full-dark treatment; Step 2.5: Perform principal component analysis on the neuroelectrical signals collected on one electrode for one round of light stimulation and one round of full-dark treatment. Perform orthogonal transformation on the high-dimensional phase points in the phase space. Arrange the features in the high-dimensional phase points in descending order of variance, and draw a trajectory plot using the first two principal components with the largest variances, so as to observe the changes in the on-chip brain dynamics trajectories under the two different states.

Images