Neuromorphic Organic Electrochemical Transistors, Neurons and Robotics in Hybrid Self-Learning Platforms

Neuromorphic platforms are emerging as new computational paradigm¹, with the ultimate goal of building artificial neural system, by replicating the efficiency of the brain in terms of parallel computation.<br/>Such technologies can mirror several aspects of the brain, by exploiting complex circuits² or by exploiting unique features of innovative materials³.<br/>Among the plethora of the available materials and structures, organic electrochemical transistors (OECTs) have emerged as ideal candidates in the building of complex neuromorphic systems, able to interface with living organisms⁴ and to learn autonomously⁵.<br/>Notably, PEDOT:PSS-based OECTs have been employed in the implementation of the first biohybrid synapse. Here, cells cultured <i>in vitro</i> could directly communicate with such artificial neurons by means of neurotransmitters, emulating a biological chemical synapse⁶.<br/>In this scenario, such devices may offer the possibility of merging electronic devices and biological neurons in an unprecedented way, addressing challenges that current technologies are not able to face, such as EMG-based prosthetics limbs failures⁷ and treatment of symptoms in neurodegenerative diseases⁸.<br/>The present work demonstrates an artificial post-synaptic neuron that drives a prosthetic hand, coordinating its grasp with the activity of pre-synaptic biological cells cultured inside the microfluidic channel of the device itself.<br/>The artificial neuron can communicate with its biological counterpart, by oxidizing neurotransmitters directly secreted by neuronal cells resulting in an update of its synaptic weight, as a stable conductive state of the polymeric channel that is then used to tune the movement of the artificial hand. In a physiological condition, such synaptic weight allows for the complete closure of the hand. Conversely, when a dysregulation of the neurotransmitter secreted by the pre-synaptic cells occurs, emulating a pathological condition, the post-synaptic neuron loses the control of the prosthetic limb. Here, a feedback controller is introduced, regulating its synaptic weight and allowing for the recovery of the physiological condition. Such result is achieved by introducing hydrogen peroxide inside the microfluidic chamber that can reverse the neurotransmitter mediated synaptic weight update. Furthermore, the neuromorphic core of such platform is exploited to achieve reinforcement learning, ultimately paving the way towards a novel generation of hybrid devices able to combine decades of development of silicon-based technologies with innovative features of organic neuromorphic devices.<br/><br/>1. Hassan, S., Humaira & Asghar, M. Limitation of Silicon Based Computation and Future Prospects. in <i>2010 Second International Conference on Communication Software and Networks</i> 559–561 (2010). doi:10.1109/ICCSN.2010.81.<br/>2. Merolla, P. A. <i>et al.</i> A million spiking-neuron integrated circuit with a scalable communication network and interface. <i>Science</i> <b>345</b>, 668–673 (2014).<br/>3. Nandakumar, S. R. <i>et al.</i> A phase-change memory model for neuromorphic computing. <i>J. Appl. Phys.</i> <b>124</b>, 152135 (2018).<br/>4. Harikesh, P. C. <i>et al.</i> Organic electrochemical neurons and synapses with ion mediated spiking. <i>Nat. Commun.</i> <b>13</b>, 901 (2022).<br/>5. Krauhausen, I. <i>et al.</i> Organic neuromorphic electronics for sensorimotor integration and learning in robotics. <i>Sci. Adv.</i> <b>7</b>, eabl5068 (2021).<br/>6. Keene, S. T. <i>et al.</i> A biohybrid synapse with neurotransmitter-mediated plasticity. <i>Nat. Mater.</i> <b>19</b>, 969–973 (2020).<br/>7. Ng, K. H., Nazari, V. & Alam, M. Can Prosthetic Hands Mimic a Healthy Human Hand? <i>Prosthesis</i> <b>3</b>, 11–23 (2021).<br/>8. Miocinovic, S., Somayajula, S., Chitnis, S. & Vitek, J. L. History, Applications, and Mechanisms of Deep Brain Stimulation. <i>JAMA Neurol.</i> <b>70</b>, 163 (2013).