Science fiction is becoming reality! Artificial neurons successfully control Venus flytraps, which may promote the transformation of brain-computer interfaces

Science fiction is becoming reality! Artificial neurons successfully control Venus flytraps, which may promote the transformation of brain-computer interfaces

Author: Cooper

In the 1980s, American computer scientist Carver Mead coined the term "neuromorphic engineering" in a paper.

He has spent more than 40 years trying to develop an analytical system that can simulate human senses and processing mechanisms (such as touch, vision, hearing, and brain thinking).

Today, while many people are still unfamiliar with what neuromorphic computing technology is, the broader technology based on the theory of these systems - artificial intelligence - is no stranger.

In recent years, the architectures of many smart chips have been influenced by neuromorphic computing research, resulting in a number of silicon architectures designed to achieve neuron-level computing capabilities and optimize computing strategies through neuromorphic computing.

Looking into the future, brain-computer interface technology is a typical neuromorphic engineering application. Scientists need to integrate artificial neuromorphic devices with biological systems to repair and enhance human functions.

However, silicon-based neuromorphic devices are greatly limited in their potential for biointegration due to their typically poor biocompatibility, complex circuits, low energy efficiency, and fundamentally different working principles from biological ion signal modulation.

Today, a new paper published in the journal Nature Communications reports an artificial neuron (OECN) that can successfully connect with the biological cells of the Venus flytrap, allowing the plant to close its leaf petals.

Figure | Neural signals collected during the experiment (Source: Nature Communications)

According to the paper, OECNs can also be integrated with all printed organic electrochemical synapses (OECSs) and respond to a variety of stimuli, defining a new prospect for local artificial neuron systems, which are expected to be integrated with the biological signaling systems of plants, insects and even vertebrates. This discovery may have implications for the future development of brain-computer interfaces and soft robotics.

The Exploration of Artificial Neurons

**It is well known that the basic building blocks of living things are fundamentally different from those of electronic devices. Therefore, the ability to connect artificial devices to biological systems is a thorny emerging field of scientific research.

Although software-based neuromorphic algorithms have been integrated into biomedical systems, hardware-based systems will ultimately be needed that are tightly coupled to living tissue and can leverage the sensing of events and the processing power of biological systems to evolve their functionality.

Neuromorphic systems, which borrow design concepts from biological signaling systems, are expected to bridge this gap.

In recent years, organic semiconductors have been highly regarded by the industry, and their applications in artificial synapses, neuroelectronics, and neural interfaces have been increasing. From a structural point of view, organic semiconductors have the characteristics of solution processability, biocompatibility, biodegradability, and softness. They can provide specific excitation, sensing, and driving capabilities, and support the transmission of electronic and ionic signals.

Artificial neurons based on organic field effect transistors (OFETs) mentioned in related reports have shown great use, but they require high voltage (5V) input to operate, which is an obvious key problem when integrating with biological organisms.

Another promising technology direction is organic electrochemical transistors (OECTs), which are gate-driven ion doping/anti-doping modulation of organic channel materials, similar to the ion-driven process and dynamics of biological systems.

Compared to OFETs, OECTs can operate at considerably lower voltages (<1 V), have higher transconductance and good threshold voltage stability, and are generally highly biocompatible, properties that make OECTs ideal candidates for developing printable, biocompatible artificial spiking neural circuits with ion-mediated spiking mechanisms that closely resemble the signaling characteristics of biological systems.

Figure | Organic electrochemical neurons and their analogy with biological neurons (Source: Nature Communications)

In this latest work, OECNs exhibit several neuronal characteristics, including ion concentration-dependent spiking and spike timing-dependent plasticity (STDP) integrated with printed organic electrochemical synapses (OECSs), which respond to a wide range of input currents (0.1–10 µA) with frequency modulation of more than 450%.

In the experiment, the researchers used the ion concentration-dependent switching properties of the transistor for the first time to regulate the spike frequency in a manner largely similar to biological systems, which is not possible in OFET-based or silicon-based neurons.

This conductance regulation enables short-term plasticity with paired-pulse facilitation and long-term plasticity that lasts for more than 1000 seconds. They anticipate that the softness of OECNs, the ability to be printed on flexible substrates, ion-modulated spikes, and multi-stimulus responses will open up new avenues for easy integration with biological neural networks.

Key links and findings

Overall, this research result has four key aspects.

First, printed organic electrochemical transistors.

The researchers chose the Axon-Hilllock (AH, referring to the region of the cell body close to the axon in a neuron) circuit to make the spiking OECN because it is the most compact model suitable for spiking neural networks (SNNs) and event-based sensors. The circuit is composed of n-type and p-type OECTs. The OECT has a lateral Ag/AgCl gate configuration with screen-printed carbon and silver electrodes on a polyethylene terephthalate (PET) substrate. The silver bottom layer reduces the line resistance when the carbon acts as a chemically inert layer in contact with the polymer semiconductor.

Second, electromechanical and chemical neurons, analogous biological neurons and biological integration.

The operating mechanism of OECNs is similar to that of biological nerve cells. In the resting state, the outside of the nerve cell has an excess positive charge and the inside of the cell has an excess negative charge, which is maintained by the insulating properties of the lipid cell membrane.

Similar to the operation of nerve cells, spikes are generated in the OECN circuit by integrating the current injected into the input terminal (Iin). Further research found that the working principle of the reset transistor is similar to the voltage-dependent potassium channel in nerve cells; membrane capacitance plays a crucial role in the conduction speed of action potentials in biological neurons. Lower membrane capacitance leads to faster propagation. In nerve cells, this reduction in capacitance is achieved by wrapping an insulating layer called myelin around the axon.

A notable feature of OECNs, especially compared to silicon- or OFET-based spiking neurons, is the ability to directly control the spike frequency by adjusting the ion concentration of the electrolyte. Similar to the leaky behavior of biological neurons that require the membrane voltage to exceed a given threshold to generate a pulse, the OECN circuit will not trigger below a certain current threshold.

Low power consumption is crucial for the application of this circuit in SNN and event-based sensors. The main power consumption source of this circuit is the amplification module. Therefore, the dynamic power consumption of the circuit is the product of IDD, dynamic and VDD. Since the inverter can operate at a low operating voltage of 0.6 V, the maximum value of IDD dynamic is 25 μA, so the maximum dynamic power consumption is 15 μW. By reducing the channel size through photolithography technology, the power consumption of OECN can be reduced to a lower value, which will reduce the current flowing through the OECT. Smaller channels will also increase the response time of the OECT and reduce energy consumption.

To demonstrate the bio-integration capabilities of the OECN, the researchers connected this fully printed neuron to a Venus flytrap. The trap closure of the flytrap can also be induced by electrical stimulation, including DC stimulation, direct charge injection, AC stimulation, and capacitive induced current, making it very suitable for integration with artificial neurons.

Third, printed organic electrochemical synapses.

OECSs are made with the same printed electrode structure as OECNs. The long-term increase in the conductivity of OECS is achieved by applying a gate voltage pulse in the channel, a process similar to the insertion of new receptors in biological synapses mediated by N-methyl-D-aspartate (NMDA) receptors, leading to a long-term increase in synaptic strength. OECS exhibited a total of 150 different states, with state retention times exceeding 1,000 seconds.

Figure | Printed organic electrochemical synapse (Source: Nature Communications)

Fourth, the integration of organic electrochemical neurons and synapses.

In biological synapses, every presynaptic input does not change the synaptic strength because this would quickly saturate the synaptic strength. The temporal correlation between presynaptic and postsynaptic neuronal spikes underlies the long-term increase/decrease in synaptic plasticity, called spike-timing-dependent plasticity (STDP), enabling associative learning.

To further illustrate the importance of OECN and OECS, this study demonstrated a simple neurosynaptic system using a synaptic transistor connected to an OECN to perform Hebbian learning (an unsupervised learning rule that closely matches how humans observe and learn about the world). Instead of inputting excitatory currents into neurons, voltage is applied to the synapse, which is converted into current based on its synaptic strength, thereby regulating the spike frequency.

The demonstration of Hebbian learning in this organic electrochemical synaptic system is an important step that can be extended to building more complex sensory and processing systems with local learning capabilities.

Figure|Sci-fi scene of the combination of man and machine in "Ghost in the Shell"

The future of “human-machine integration” is promising

The researchers said that compared with silicon-based circuits, OECN neural synaptic systems with STDP have far fewer components, the circuits can be printed on a large scale and have high manufacturing capacity, and compared with OFET-based circuits, neurons can be completely printed on flexible substrates and operate at lower power, so distributed low-cost intelligent units can be developed for the future Internet of Things.

The spike frequency of OECN can be adjusted by changing the input current, membrane capacitance and voltage of the amplifier. This property, as well as the ability to adjust the peak frequency by adjusting the electrolyte concentration, makes it easier to integrate with biological systems and is expected to promote the development of future implantable devices.

The researchers demonstrated this possibility by connecting OECN to a Venus flytrap and inducing its lobe closure based on the firing frequency of the neuron. In the future, the unique ability to sense a variety of biological, physical, and chemical signals will enable a variety of sensor detection.

Figure|Super biological computer in the animation "EVA"

Looking to the future, the possibility of fusing multiple sensing elements into the neuron itself could enable the development of new types of bio-integratable sensors with applications ranging from smart neuromorphic packaging for the Internet of Things, continuous body health monitoring (i.e., wearable electronics), to brain-computer interfaces.

The local artificial neural synaptosome system composed of OECNs and OECSs is expected to generate more possibilities by integrating with the signaling system of plants, the diffuse nervous system of invertebrates and vertebrates, the peripheral nervous system, and the central nervous system.

In many science fiction works, humans have boldly envisioned the perfect combination of organisms and machines in the future. Although it is very cyberpunk, it is not groundless. Perhaps with the advancement of technology, when scientists are able to overcome the various challenges of human-machine integration, the dream will come true.

References:

https://www.nature.com/articles/s41467-022-28483-6

Source: Academic Headlines

<<:  Where does the smell of cilantro come from? Is your love for it determined by your genes?

>>:  Why are girls’ abdominal muscles called vest lines?

Recommend

12 Best Strategies for Attracting and Retaining App Users

As more and more companies create independently b...

Why can’t mobile phones be multi-tasked like computers?

The human brain seems to be naturally equipped wi...

Case Study: 12 Gamification Strategies for User Growth

This article uses a case study to introduce how t...

Mate7 is popular, so Huawei's problem comes

It is no exaggeration to say that the emergence o...

Unity Awards 2015 is about to start, and good games are coming soon!

Unity Awards 2015 (Game and Application Creativit...