Bioinspired variable encoding neurocomputing chip and method of making same

Abstract

The application discloses a variable coding neural computing chip based on biological intelligence and a preparation method thereof, and relates to cell culture technology and sensor technology. The neural computing chip is composed of two-layer sub-chip systems. The first-layer sub-chip system is a neural electronic interface based on micro-electro-mechanical system technology, mainly including microelectrodes of various geometric structures modified with platinum-iridium nanomaterials, lead wires and macroelectrodes as contact points; the second-layer sub-chip system is an active neural network using ex vivo cultured nerve cells to provide biological intelligence. The chip provides various coding electric stimulations through the microelectrodes of various geometric structures, increases the operation range of neural computing, improves the chip performance through the platinum-iridium nanomaterials, prolongs the effective working time of the chip, and realizes the neural computing chip with biological intelligence through the joint work of the two-layer sub-chip systems.

Description

TECHNICAL FIELD

[0001] The present application relates to the fields of biosensors, micro-electro-mechanical systems, neural cell culture, brain-computer interface, neural computing, etc., and particularly relates to a variable coding neural computing chip based on biological intelligence and a preparation method thereof. BACKGROUND

[0002] Neural computing using biological natural intelligence is the most promising technology beyond artificial intelligence, which has significant advantages in information processing efficiency, such as speed and power consumption. The core lies in that neurons can efficiently transmit information in the form of electrical pulses, and can generate a complex network of hyperlinks to carry out large-scale coordinated parallel processes. There are great ethical and moral controversies in the application of biological intelligence based on in-vivo brain-computer interface, while the in-vitro neural network obtained by cell culture technology also has biological intelligence, and has the advantages of good consistency and stability, and is easy to integrate with the neural electronic interface based on microelectrode array to form a neural computing chip. The neural computing chip based on in-vitro cell culture neural network has attracted more and more attention. At present, the neural computing chip with biological intelligence prepared based on in-vitro cultured neural network has realized the functions of controlling robots to avoid obstacles and walk in the maze, and even playing ping-pong games in a virtual game world. However, the existing neural computing chip has the problems of single coding mode and short effective working time. SUMMARY

[0003] In order to solve the above problems and effectively utilize the natural neural network with biological activity, the present application optimizes the design and preparation of the neural computing chip, and strives to realize various spatial coding of electrical stimulation regulation by combining the diversity and combination ability of the geometric morphology of the electrode, so as to non-destructively regulate and intervene the neural activity; and to realize the processing of the micrometer-level structure and nanometer-level characteristics of the neural chip by combining the electrochemical surface modification, so as to improve the working performance of the neural chip, improve the connection between the neural-electronic interface, and prolong the working time.

[0004] The purpose of the present application is to prepare a neural chip with plasticity and low cost and high computing power. In view of this, the present application provides a variable coding neural computing chip based on biological intelligence and a preparation method thereof, so as to at least partially solve at least one of the above-mentioned technical problems, and to realize parallel and efficient neural computing by using biological intelligence.

[0005] In order to achieve the above-mentioned purpose, the main technical scheme of the present application includes:

[0006] As one aspect of the present application, a variable coding neural computing chip based on biological intelligence is provided, which includes a two-layer sub-chip system.

[0007] The first layer sub-chip system is a neural electronic interface for bidirectional interaction with the neural network, and the second layer sub-chip system is an active neural cell network with biological intelligence. The first layer sub-chip can interact with the second layer sub-chip to realize signal input and readout.

[0008] The neural electronic interface has a microelectrode array capable of supporting close adhesion and growth of neural cells and a macroelectrode capable of matching external detection circuitry. The neural electronic interface has a neural electronic interface. The neural electronic interface substrate is a quartz glass with an area of 5*5 cm 2 , a thickness of 1 mm; a composite metal layer with a sandwich structure of chromium / platinum / chromium (30 / 250 / 10 nm) as a conductive metal layer; SiO2 with a thickness of 800 nm as an insulating layer; and the surface is decorated with platinum-iridium nanoparticles of nanometer size (<100 nm).

[0009] The microelectrode array includes 59 bidirectional microelectrodes of various geometric shapes and 1 reference electrode, and the arrangement presents a double circular arc shape with central symmetry. The microelectrode site feature size is 10-70 μm (for example, the diameter of a circular microelectrode site is 30 μm, the long side of a rectangular microelectrode site is 70 μm, the short side is 10 μm, and the side length of a square microelectrode site is 25 μm). The site area is 700±80 μm 2 , and the total coverage area of the microelectrode arrangement is 11.56 mm 2 , arranged in a double circular arc shape with central symmetry in the shape of X, capable of providing 6 different electric field spatial distributions, a total of 59 channel unit point electric stimulation modes, and more than 3000 different electric field spatial distribution double-site electric stimulation modes. The microelectrode array material is chromium / platinum (30 / 250 nm), and the surface is decorated with platinum-iridium nanocomposite nanomaterial (<100 nm). The size of the macroelectrode is 5*5 mm 2 , used as a contact to connect with an external circuit board, and the macroelectrode material is chromium / platinum (30 / 250 nm).

[0010] The active neural cell network is derived from an in vitro cultured neural cell of a mouse or a rat, and needs to be cultured on the neural electronic interface to have a super node and super link network structure with more than 5000 nodes and more than 10000 links and generate detectable electrophysiological signals. As a preferred embodiment, the in vitro cultured neural cell includes a hippocampal neural cell or a cortical neural cell.

[0011] As another aspect of the present application, a preparation method of a variable coding neural computing chip based on biological intelligence is also provided, including the following preparation steps:

[0012] The preparation of the first layer sub-chip system comprises: a. substrate cleaning, b. glue spinning lithography, c. metal layer sputtering, d. metal layer stripping, e. plasma enhanced chemical vapor deposition insulation layer, f. plasma etching to expose microelectrode and macroelectrode sites, g. wet etching of the chromium metal layer of the microelectrode and macroelectrode sites, h. sticking a glass ring on the microelectrode array to manufacture a cell culture chamber, and i. electrochemical surface modification of platinum-iridium nanocomposite nanomaterial;

[0013] The integration of the second layer sub-chip system and the first layer sub-chip system comprises: a. cleaning the neural electronic interface, b. inoculating a neural cell suspension taken from active brain tissue, dissociated and dispersed, into the cell culture chamber on the microelectrode array, and c. culturing in a suitable sterile environment to form a functional neural network.

[0014] Based on the above technical solutions, the present application has at least one or part of the following beneficial effects compared with the prior art:

[0015] (1) The in vitro cultured neural network is used to complete the neural computing function, has high controllability, consistency, and wide application scenarios and commercial prospects;

[0016] (2) The neural electronic interface modified with platinum-iridium nanomaterial is used to interact with active neural cell networks with biological intelligence, has high spatial and temporal resolution, high-speed real-time bidirectional regulation advantages, and realizes fast information exchange with the neural network;

[0017] (3) A variable coding neural computing chip based on biological intelligence has high biocompatibility, can support the growth and functional operation of active neural cell networks, and can also realize the injection of various coded electric stimulation signals, thereby increasing the operation range of neural computing. It helps to break through the limitation of artificial intelligence operation capacity and realizes low-power large-scale parallel operation. BRIEF DESCRIPTION OF DRAWINGS

[0018] Figure 1 It is a structure schematic diagram of the variable coding neural computing chip based on biological intelligence.

[0019] Figure 2 It is a structure schematic diagram of the first layer sub-chip system of the present application.

[0020] Figure 3 It is a local structure schematic diagram of the microelectrode site area of the first layer sub-chip system of the present application.

[0021] Figure 4 It is a working principle diagram of the neural computing chip of the present application.

[0022] Figure 5 It is a flowchart of the preparation process of the main components of the first layer sub-chip system of the present application.

[0023] Figure 5 a. Spin AZ500 photoresist on the glass substrate after piranha solution cleaning and pre-bake curing;

[0024] Figure 5 b. Patterned exposure and development of the photoresist using custom mask;

[0025] Figure 5 c. Sputtering of chromium / platinum / chromium (30 / 250 / 10 nm) conductive metal layer on the glass substrate after the first lithography;

[0026] Figure 5 d. Lift-off of the device after sputtering of the metal layer, leaving the microelectrode, macroelectrode and lead;

[0027] Figure 5 e. PECVD (plasma enhanced chemical vapor deposition) deposition of silicon oxide (800 nm) insulating layer on the device after lift-off and cleaning;

[0028] Figure 5 f. Spin AZ500 photoresist on the device after deposition of the insulating layer and pre-bake curing;

[0029] Figure 5 g. Patterned exposure and development of the photoresist using custom mask;

[0030] Figure 5 h. RIE (reactive ion etching) removal of the insulating layer at the locations without photoresist protection, achieving window exposure of the microelectrode and macroelectrode sites;

[0031] Figure 6 Flowchart of two-layer sub-chip system integration in the preparation process of the neural computing chip of the present application

[0032] In the above figures, the meanings of the reference signs are as follows:

[0033] 1. First layer sub-chip system: microelectrode array as neural electronic interface; 2. Second layer sub-chip system: active neural cell network providing biological intelligence; 3. Counter / reference electrode; 4. Glass substrate; 5. Macroelectrode: as contact; 6. Microelectrode; 7. Lead. DETAILED DESCRIPTION

[0034] The application provides a variable coding neural computing chip based on biological intelligence and a preparation method thereof. The chip realizes utilization of biological intelligence through a neural electronic interface with variable electric stimulation coding capability, thereby realizing multi-input parallel decision operation. The chip provides various spatial distribution electric fields through microelectrodes with various geometric structures, thereby performing variable coding, increasing the operation range of neural computing, improving the performance of the chip through platinum-iridium nanomaterials, prolonging the effective working time of the chip, and realizing the neural computing chip with biological intelligence through joint working of two-layer sub-chip systems.

[0035] The operation decision function of the brain has strong application prospects due to low power consumption and parallel processing capability. In order to realize full utilization of the brain within the ethical and moral norms, the ex vivo cultured neural network is gradually valued. The ex vivo cultured neural network has biological intelligence and can perform high-level functions such as learning and memory, and is confirmed in use scenarios such as playing games and controlling robots. In order to more conveniently use the biological intelligence, the active biological neural network is packaged with a microelectrode array as a neural electronic interface, so that various electric stimulation inputs can be used to realize information injection and information reading out through decoding electrophysiological detection signals.

[0036] In order to make the purpose, technical scheme and advantages of the application clearer and more apparent, the application is further described in detail below in combination with specific examples and with reference to the drawings. However, it should be noted that the following examples are only used to illustrate the technical scheme of the application, but the application is not limited thereto.

[0037] Figure 1 The variable coding neural computing chip based on biological intelligence is shown in the structural schematic diagram. It includes two-layer sub-chip systems.

[0038] The first-layer sub-chip system 1 is a microelectrode array as a neural electronic interface, and the second-layer sub-chip system 2 is an active neural cell network with biological intelligence. The first-layer sub-chip system 1 can interact with the second-layer sub-chip system 2 to realize signal input and reading out. The second-layer sub-chip system 2 can convert signals and complete information operation processing.

[0039] Figure 2 The structure schematic diagram of the first-layer sub-chip system of the application is shown. The first-layer sub-chip system 1 includes a counter / reference electrode 3, a glass substrate 4, a macroelectrode 5 as a contact, a microelectrode 6, a lead 7 and the like. The counter / reference electrode 3, the macroelectrode 5, the microelectrode 6 and the lead 7 are processed on the glass substrate 4. The function of the lead 7 is to connect the counter / reference electrode 3 and the macroelectrode 5, so as to connect an external circuit system to apply an electric stimulation signal and collect an electrophysiological signal. The length x width of the glass substrate 4 is 5 x 5 cm 2 , and the thickness is about 1 mm.

[0040] Figure 3 The microelectrode site area local structure diagram of the first layer sub-chip system of the application is shown. The microelectrode array includes 59 bidirectional microelectrodes of different geometric shapes and one larger area counter / reference electrode 3, and the bidirectional microelectrodes of different geometric shapes are arranged in a central symmetric double circular arc shape, with a characteristic size of 10-70 μm, and a site area of 700±80 μm 2 . The specific size is shown in the figure, the radius of the circular electrode is 15 μm; the side length of the square electrode is 25 μm; the side length of the equilateral triangle electrode is 40 μm; the inner diameter of the semi-circular ring electrode is 20 μm; the outer diameter is 30 μm; the circular ring width is 10 μm; the length x width of the rectangular electrode is 70 x 10 μm 2 . The total coverage area of the microelectrode arrangement is 11.56 mm 2 , which is in a central symmetric double circular arc shape in the form of X, can provide 6 different electric field spatial distributions, a total of 59 channel unit point electric stimulation modes, and more than 3000 different electric field spatial distribution double site electric stimulation modes.

[0041] Figure 4 The working principle diagram of the neural computing chip of the application is shown. Specifically, the electric stimulation pulse sequence is applied to the first layer sub-chip system 1 through selected electrodes. Different microelectrodes, microelectrode pairs, and microelectrode groups can generate different distributed electric fields and have different effects on biological neural networks. A single electric pulse stimulation sequence can realize a variety of encoded stimulation paradigms by changing the application position. The first layer sub-chip system 1 directly interacts with the second layer sub-chip system 2 through microelectrodes. The active neural cell network of the second layer sub-chip system 2 can be divided into an input layer, a hidden layer, and an output layer: the input layer is a group of nerve cells in close contact with the microelectrode and directly receiving external stimulation regulation, and various stimulation signals are input to the hidden layer for processing after biological signal conversion by the input layer cells; the hidden layer is an overall biological neural network with super-multiple nodes and super-multiple links; the output layer converts the biological signal to the electronic signal. The super-multiple nodes and super-multiple links refer to more than 5000 nodes and more than 10000 links. The layers can be multiplexed. The signal processed by the second layer sub-chip system 2 is read out by the first layer sub-chip system 1. By analyzing the changes of the action potential / field potential of the neural signal, the processing result of the biological neural network on the electric stimulation pulse sequence can be decoded.

[0042] Figure 5 The flow chart of the preparation process of the main components of the first layer sub-chip system of the application is shown. The steps include:

[0043] (a) Clean the glass substrate with Piranha solution for 3 times to make sure there is no organic residue on the glass substrate. After baking dry, spin AZ500 photoresist on the glass substrate and pre-bake, the thickness is about 2.5 μm, other positive photoresist can also achieve similar results;

[0044] (b) Expose the photoresist on the glass substrate with custom mask and develop, thus the pattern as shown in Fig. 1 is fabricated on the glass substrate; Figure 2

[0045] (c) Sputter Cr / Pt / Cr (30 / 250 / 10 nm) conductive metal layer on the glass substrate after the first lithography, the first layer of Cr is used to increase the adhesion between the Pt metal layer and the glass substrate, the second layer of Cr is used to increase the adhesion between the Pt metal layer and the SiO2 insulating layer, to reduce the stress difference between the layers of materials;

[0046] (d) Lift-off after sputtering the metal layer, thus only the metal microelectrode, macroelectrode and lead are left;

[0047] (e) PECVD (Plasma Enhanced Chemical Vapor Deposition) deposit SiO2 (800 nm) insulating layer after lift-off and cleaning, low temperature chemical vapor deposition can also achieve similar results;

[0048] (f) Spin AZ500 photoresist on the device after depositing the insulating layer and pre-bake, this layer of photoresist is used to protect the part of the lead that does not need to be exposed by windowing, and to fabricate the shape of the electrode site on the device;

[0049] (g) Expose the photoresist with custom mask and develop, so that the subsequent non-selective RIE (Reactive Ion Etching) can selectively remove the insulating layer at the site of discharge;

[0050] (h) Remove the insulating layer at the position without photoresist protection by RIE (Reactive Ion Etching), to achieve the windowing exposure of the microelectrode and macroelectrode sites, and use a multimeter to check whether the site surface can conduct.

[0051] After completing these steps, wet etching is also needed, to selectively remove the Cr metal layer at the microelectrode and macroelectrode sites with Cr etching solution to ensure that the sites have lower impedance. After thoroughly cleaning the microelectrode array, a glass ring is attached to the microelectrode array to manufacture a cell culture chamber. Electrochemical surface modification is also needed in this embodiment, to fabricate a layer of platinum-iridium nanocomposite nanomaterial on the microelectrode site to ensure better biocompatibility and higher quality of neural signals.

[0052] Figure 6 ​The flow chart of two-layer sub-chip system integration in the preparation process of the neural computing chip of the present application is shown in Figure 1. First, the first layer of sub-chip system 1 needs to be sterilized by alcohol and ultraviolet light. Then the electrode surface needs to be coated with a chemical substance that is conducive to cell growth, and in this embodiment, polylysine is used. Finally, the prepared neural cell suspension is inoculated into the cell culture chamber and cultured under suitable conditions until a functional neural network is formed on the microelectrode surface.

[0053] Example 1: Processing of parallel input electrical stimulation pulse sequence using the above variable coding neural computing chip.

[0054] (1) Installation of bidirectional communication system. Connect the neural computing chip to the interface amplification circuit, and then to the electrophysiological recording instrument. In this embodiment, the electrophysiological recording instrument used is the Cerebus electrophysiological signal detection instrument. The leads of the electrical stimulator used to apply the electrical stimulation pulse sequence are connected to the ground and the selected stimulation site, respectively. In this embodiment, the electrical stimulator used is the Multichannel electrical stimulation instrument.

[0055] (2) Making a single stimulation-response table of decoded neural activity. First, record the spontaneous neural activity of the neural computing chip as the baseline of the chip. Then, apply multiple electrical stimulation pulse sequences to the single site of the neural computing chip in succession, and by comparing the induced neural activity with the spontaneous activity, obtain the stimulation-response correspondence table of multiple single stimuli.

[0056] (3) Processing of parallel information. Apply electrical stimulation pulse sequences to two sites of the neural computing chip, respectively, and by comparing the induced neural activity with the spontaneous activity, obtain the processing result of the neural computing chip on the parallel input electrical stimulation pulse sequence. By jointly analyzing the single stimulation-response table and the processing result of parallel input, the internal hidden layer neural network structure and processing paradigm can be deduced in reverse, and thus the unique computing ability of the neural computing chip can be utilized.

[0057] The above specific embodiments further illustrate the purpose, technical solutions and beneficial effects of the present application. It should be understood that the above description is only a specific embodiment of the present application and is not intended to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A bio-smart based variable encoding neurocomputing chip, characterized in that, The chip includes two sub-chip systems: The first sub-chip system is a neural electronic interface for bidirectional interaction with a neural network; The second sub-chip system is an active neural cell network with biological intelligence; The first sub-chip system can interact with the second sub-chip system to realize signal input and readout; The neural electronic interface includes: A microelectrode array that can support close adhesion and growth of the neural cells; A macroelectrode that can match external detection circuitry; In the neuroelectronic interface, the thickness of the substrate of the neuroelectronic interface is not less than 5x5 cm in size of 1 mm 2 of a quartz glass sheet; The neural electronic interface uses a composite metal layer with a chromium-platinum-chromium sandwich structure as a conductive metal layer, the first layer of chromium has a thickness of 30-50 nm, the second layer of platinum has a thickness of 200-300 nm, and the third layer of chromium has a thickness of 5-15 nm; The neural electronic interface uses SiO2 with a thickness of 600-800 nm as an insulating layer; The surface of the neural electronic interface is modified with nano-sized platinum-iridium nanoparticles; The microelectrode array includes 59 bidirectional microelectrodes of various geometric shapes and 1 reference electrode, and the arrangement presents a double-arc shape with central symmetry: The microelectrode has an electrode site feature size of 10-70 μm and an electrode site area of 700 80 μm 2 ; The total covering area of the arrangement of the microelectrodes is 11.56 mm 2 X-shaped in a central symmetrical double circular arc shape; The microelectrode array can provide 6 different electric field spatial distributions, a total of 59 channel unit point electric stimulation modes, and more than 3000 different electric field spatial distribution double-site electric stimulation modes; The material of the microelectrode array is chromium, platinum, and platinum-iridium composite nanomaterial.

2. The chip of claim 1, wherein: In the macroelectrode, The macroelectrode has dimensions of 5 x 5 mm 2 and is made of chromium, platinum.

3. The chip of claim 1, wherein: The active neural cell network is derived from an in vitro cultured neural cell of a mouse or a rat; The active neural cell network needs to be cultured on the neural electronic interface to have a super node and super link network structure with more than 5000 nodes and more than 10000 links, and to generate detectable electrophysiological signals; The active neural cell network is divided into an input layer, a hidden layer, and an output layer: the input layer is a group of neural cells in close contact with the microelectrode that directly receives external stimulation regulation, and various stimulation signals are input to the hidden layer after biological signal conversion by the input layer cells; the hidden layer is an integrated biological neural network with super node and super link; the output layer converts the biological signal of the hidden layer calculation result to an electronic signal for easy readout; The in vitro cultured neural cells include hippocampal neural cells or cortical neural cells.

4. The chip of claim 1, wherein Various applied electric stimulation modes are used as encoding signal inputs, and the detected neural electrophysiological signals are used as calculation result outputs; The applied electric stimulation modes include frequency encoding, timing encoding, spatial position encoding, and non-linear encoding, and the detected neural electrophysiological signals include discontinuous neural pulse discharge signals and continuous field potential signals.

5. A preparation method of the variable encoding neural computing chip based on biological intelligence according to claim 1, characterized in that: The preparation of the first layer sub-chip system comprises the following steps: a. substrate cleaning, b. glue spinning lithography to pattern the microelectrode array, c. metal layer sputtering, d. metal layer stripping, e. plasma enhanced chemical vapor deposition of the insulating layer, f. plasma etching to expose the electrode sites, g. wet etching of the top chromium layer of the electrode sites with chromium etching solution, h. sticking glass ring to make cell culture chamber, i. electrochemical surface modification of platinum-iridium nanocomposite nanomaterials; The integration of the second layer sub-chip system and the first layer sub-chip system comprises the following steps: a. cleaning the neural electronic interface, b. inoculating the neural cell suspension taken from active brain tissue, dissociated and dispersed into the cell culture chamber on the microelectrode array, c. culturing in a sterile environment to form a functional neural network.

Images