Max Plank AI Researchers Have Developed Bio-Realistic Artificial Neurons That Can Work In A Biological Environment And Can Produce Diverse Spiking Dynamics

The development of neuromorphic electronics depends on the effective mimic of neurons. But artificial neurons aren’t capable of operating in biological environments. Organic artificial neurons that work based on conventional circuit oscillators have been created, which require many elements for their implementation. An organic artificial neuron based on a compact nonlinear electrochemical element has been reported. This artificial neuron is sensitive to the concentration of biological species in its surroundings and can also operate in a liquid. The system offers in-situ operation, spiking behavior, and ion specificity in biologically relevant conditions, including normal physiological and pathological concentration ranges. While variations in ionic and biomolecular concentrations regulate the neuronal excitability, small-amplitude oscillations and noise in the electrolytic medium alter the dynamics of the neuron. A biohybrid interface is created in which an artificial neuron functions synergistically with biological membranes and epithelial cells in real-time.

Neurons are the basic units of the nervous system that are used to transmit and process electrochemical signals. They operate in a liquid electrolytic medium and communicate via gaps between the axon of presynaptic neurons and the dendrite of postsynaptic neurons. For effective brain-inspired computing, neuromorphic computing leverages hardware-based solutions that imitate the behavior of synapses and neurons. Neuron like dynamics can be established with conventional microelectronics by using oscillatory circuit topologies to mimic neuronal behaviors. However, these approaches can mimic only specific aspects of neuronal behavior by integrating many transistors and passive electronic components, resulting in a bulky biomemtic circuit unsuitable for direct in situ biointerfacing. Volatile and nonlinear devices based on spin torque oscillators or memristor can increase the integration density and emulate neuronal dynamics. Memristive device based artificial neurons have the potential for high integration density because of their negative differential resistance. 

Spin torque oscillators are magnetic nanodevices that work well with silicon technology, and their nonlinearity and dynamics have recently been leveraged for their audible source recognition and spoken language. But there is no feasible route for biointerfacing with spin torque oscillators as the oscillatory dynamics of these are too fast in real-time, and they demand a magnetic environment as well. Other approaches for spiking and circuits have been developed, but they were also found to encounter similar problems making artificial neurons insufficiently capable. Organic electrochemical devices offer an alternative to neuromorphic electronics. These electronics can operate in a closer environment to biology because of their soft nature and capacity to interact with ions in aqueous electrolytes directly. The article’s primary focus is to report an organic artificial spiking neuron, developed by Max Plank’s researchers, that has electrobiochemical degrees of control and enables in situ neuromorphic sensing and biointerfacing. The researchers further created a biohybrid interface in which the artificial neurons function synergistically with the membranes of epithelial cells.

Understanding The Organic Artificial spiking Neuron

The Organic Artificial Neuron (OAN) consists of compact nonlinear building blocks made of two organic electrochemical transistors (OECTs) T1 and T2, both of which are p-type transistors with T1 being a depletion mode transistor and T2 being an enhancement mode transistor. These OECTs primarily operate in an aqueous environment and are sensitive to ionic and polyatomic ions. Both the gate and channel of an OECT are in direct contact with the electrolyte. Ion-electronic charge compensation results in a large current modulation. This high gate voltage produces a high transconductance to drain current modulation, the hallmark of OECTs. Another essential attribute of OECTs is the dependency of ID on the ion concentration of the electrolyte. To be more specific, the mobile ions provided by the electrolyte electrostatically balance the fixed charges of the polymeric channel’s ion-conducting phase. This Donnan equilibrium results in a voltage drop at the electrolyte/polymer interface, which is mirrored by OECT threshold voltage. When the organic electrochemical nonlinear device (OEND) is coupled to an RC element to form an OAN, its response is bifurcated, resulting in voltage or current oscillations. Ions can enter the polymer’s core in OECTs, and the volumetric nanoscale . The action potentials of OAN are represented by these spike-based oscillations. The amplitude and window of voltage/current oscillations can be precisely designed by engineering the threshold voltages T1 and T2. The OAN displays the salient features noticed in the biological neurons spiking response. The tendency of the neuron to fire spikes in the OAN can be modulated by electrochemical oscillations and transmitted by ionic fluxes. Because of its finite response time, the OAN has a stimulus-response delay and behaves as a temporal integrator. Recurrence plots of the spike-to-spike amplitudes and interspike intervals show that the coding scheme changed from tonic to noise-induced bursting activity and that spiking was resilient to noise in the electrolytic media.

In situ spike-based neuromorphic sensing- Concept

According to estimates, the extracellular electrolytic space makes up between 15 and 30 percent of the brain’s total volume. Different ionic species. It maintains a homeostatic balance of ion concentration. But these ion concentrations can be changed in different spatiotemporal scales. Due to its liquid electrolytic operation, the OAN exhibits a firing property dependent on the host electrolyte. For the OAN to operate in situ with biological membranes and neurons, it must be sensitive at biophysically relevant ionic concentration ranges. Any change in the ionic concentration gradient between the intracellular and extracellular medium results in an alteration of threshold/excitability of the biological neuron, and firing can also be initiated by varying these concentrations. As a modulatory neurotransmitter, dopamine controls a number of crucial mental processes, including motivation, learning, motor control, mood regulation, and addiction. At the cellular level, this can impact neuronal excitability in different ways via synaptic and non-synaptic activation of dopamine receptors. Neuronal signaling is characterized by the passage of inward and outward ionic currents across the ion channels of biological membranes. Numerous channelopathies, which can result in significant clinical disorders, including cystic fibrosis and myotonia congenita, have the effect of dysregulating these pathways. The OAN response gains additional biophysical realism by incorporating on-chip selectivity and specificity traits similar to biological ion channels.

Details- Biohybrid Neuron

A biohybrid neuron is created by incorporating a biohybrid neuron between the gate and the channel of T1 and consists of a biological and artificial compartment. This neuron functions both in situ and in real-time. These biohybrid systems can be used as manipulable in vitro models for fundamental research, such as to comprehend the underlying mechanisms of neuronal signaling, as well as a platform for investigating the physiopathological conditions that affect the barrier integrity of biological tissues or the impact of extraneous physicochemical cues. A relatable example of the biomembrane model system is the prototype of epithelial cell line Caco-2, a model of the intestinal epithelial barrier incorporated with OAN and used for in vitro toxicology and drug delivery studies. Also, it is to be noted that it is necessary to have similar dimensions in both the domains for interfacing of OAN with biological neurons. Because the device dimensions are so crucial to the OAN’s spiking response, careful design of the OAN is required.

A nonlinear electrochemical-based OAN has been reported that is inspired by the properties of the biological neurons functioning the wet surroundings. Nonlinear phenomena is exhibited by artificial neurons that depend on the composition of biophysical host environments. OAN can simulate The aqueous electrolyte that makes up this extracellular region contains the biological sensitivity to ionic and biomolecular species in a nearby aquatic environment. A biohybrid interface using OANs was modulated by using biological membranes of epithelial cells in real-time and in situ. Also, a comparison of the state-of-the-art technology has been provided. The noise, electrolytic potential, and the regional concentrations of particular ionic and biomolecular species all affect neurons’ excitability, dynamics, and spiking characteristics. OAN requires only two transistors, Unlike typical organic ring oscillators, which have many transistors, making the OAN capable enough to merge into a single device. Whereas, for practical applications, the variability of soft matter devices and integrability should be further developed. Also, bio fuelled, and self-sustainable oscillators should be used rather than powering the OAN externally to emulate certain metabolic pathways of biological neurons. Last but not least, latent “memory time windows” in the context of dopamine detection can serve as the foundation for on-chip learning phenomena like biomolecular reward prediction error coding.

This Article is written as a research summary article by Marktechpost Staff based on the research paper 'An organic artificial spiking neuron for in situ neuromorphic sensing and biointerfacing'. All Credit For This Research Goes To Researchers on This Project. Check out the paper and reference article.
Please Don't Forget To Join Our ML Subreddit
­čÉŁ Join the Fastest Growing AI Research Newsletter Read by Researchers from Google + NVIDIA + Meta + Stanford + MIT + Microsoft and many others...