Лаборатория стохастических
мультистабильных систем

Laboratory of Stochastic
Multistable Systems

StoLab employees proposed the concept of a memristive neurohybrid chip


StoLab employees proposed the concept of a memristive neurohybrid chip for use in compact biosensors and neuroprostheses

Scientists at Lobachevsky University, in collaboration with colleagues from Russia, Italy, China and the United States, were the first to propose the concept of a memristive neurohybrid chip for use in compact biosensors and neuroprostheses. The corresponding expert opinion in the form of a perspective article was published in the journal Frontiers in

Neuroscience (https://doi.org/10.3389/fnins.2020.00358) as part of the special issue "New technologies and systems for biologically plausible implementation of neural functions."

The concept is based on existing and promising solutions at the junction of neural cellular and microfluidic technologies that allow the growth of a spatially ordered living neural network, in combination with CMOS-compatible technologies for creating microelectrode arrays and arrays of memristive devices for recording, processing and stimulating bioelectric activity in real time.

The interaction of different subsystems is organized on one crystal (chip) and is controlled by built-in analog-digital circuits. The implementation of such a biocompatible microelectronic system, along with the development of cellular technologies, will provide a breakthrough in the field of neuroprosthetics with an important competitive advantage: a miniature bioelectric activity sensor based on micro- and nanostructures with the ability to store and process signals in both forward propagation and feedback modes will play the role of an active neurointerface for intelligent control of neural structures. Such possibilities are unattainable on the basis of traditional architectures of neurointerfaces and can be extended to other types of bioelectric signals to solve the problems of recording signals of activity of the brain, heart and muscles, as well as the state of the skin as part of wearable signal processing and diagnostics systems.

Currently, for the development and creation of bidirectional neurointerfaces, complex electronic circuits are used that implement special mathematical models and neuromorphic principles of information processing. Such electronic systems are based on traditional components and do not meet the requirements for energy efficiency and compactness for safe interaction with living cultures or tissues on a single chip.

Memristor has a unique property of nonlinear resistive memory and is a promising element of analog information processing systems, including those with a neuron-like structure, and can also serve as a sensor of electrophysiological activity with the function of simultaneous accumulation and non-volatile storage of information.

According to the authors of the article, it is brain-like memristive systems that will be able to provide the highest degree of adaptability, energy efficiency, and scalability required for the implementation of compact and efficient neural interfaces.

The proposed neurohybrid system is shown schematically in Figure 1A and consists of several functional layers integrated on a single CMOS chip. The upper layer is a part of the neuronal system, represented here by a culture of dissociated hippocampal cells grown on a multi-electrode array and functionally ordered using a special microfluidic channel layout shown in Figure 1B.

The microelectrode layer serves for extracellular registration and stimulation of neurons in vitro and is implemented in the upper metallization layer of the CMOS device layer together with an array of memristive devices (Figure 1D).
The simplest task performed by memristive devices is the direct processing of spike activity of a biological network (Figure 1C), however, promising neural network architectures based on fully connected memristive arrays "cross-bar" with the possibility of self-learning are designed for adaptive decoding of temporal and spatial characteristics of bioelectric activity.

The outputs of such a network (Figure 1F) can be used to control cell exposure by sequential modulating of extracellular stimulation according to a given protocol (Figure 1G).

Analog and digital circuits for controlling arrays of electrodes and memristive devices, amplification, generation and transmission of signals between layers should be implemented in the main CMOS device layer (Figure 1E).

To create a neurohybrid chip will require the joint design and optimization of all these elements at the levels of materials, devices, architectures and systems. Of course, this work should keep pace with the development of bio- and neurotechnologies to solve a number of problems discussed in detail in the article and related primarily to biocompatibility, mechanical action, geometry, arrangement and miniaturization of microelectrodes and probes, as well as the reaction of living culture / tussue on the interface with an artificial electronic subsystem.

The concept reveals the idea of ​​creating a "brain-on-a-chip" system belonging to a more general class of memristive neurohybrid systems for next-generation robotics, artificial intelligence and personalized medicine.
To illustrate the proposed approaches and related products in a foreseeable timeline, a roadmap of memristive neuromorphic and neurohybrid systems is proposed (Figure 2).
The key direction of development in it is associated with the development and making (serial production) of specialized software based on the architecture and mechanism of functioning of biological neural networks to support the development and mass implementation of artificial intelligence technologies, machine learning, neuroprosthetics and neurointerfaces.
The roadmap tentatively begins in 2008 with the current wave of interest in memristors and includes ongoing research and development in the broad areas of neurobiology and neurophysiology.

The following product niches in the roadmap at different stages of development in this area:

  • neuromorphic computing devices;
  • non-invasive neurointerfaces;
  • neuroimplants, neuroprostheses and invasive neurointerfaces.

It is the unique memristive devices that determine their crucial importance in the development of applied neuromorphic and neurohybrid systems for the properties of neurocomputing devices, brain-computer interfaces, and neuroprosthetics.

These spheres will occupy a significant part of the global hi-tech market of trillions of dollars by 2030, taking into account the speed of development and implementation of artificial intelligence technologies, the Internet of things, technologies of "big data", "smart cities", robotics, and in the near future - also neuroprosthetics and instrumental adjustment / support / enhancement of human cognitive abilities.

Various research tasks within the framework of the concept are already being actively solved at UNN with the support of the Science Foundation (grant No. 16-19-00144) - in part of creating arrays of metal-oxide memristors for a bidirectional neurointerface, the Russian Foundation for Basic Research (grants No. 18-29 -23001 and 20-01-00368) - in part of brain-like memristive neural network architectures and spike neural networks, as well as the Government of the Russian Federation (Agreement No. 074-02-2018-330 (2)) - in part of noise-induced phenomena in memristive materials, devices and networks.