Automatic Adaptation of Model Neurons and Connections to Build Hybrid Circuits with Living Networks

Abstract

Hybrid circuits built by creating mono- or bi-directional interactions among living cells and model neurons and synapses are an effective way to study neuron, synaptic and neural network dynamics. However, hybrid circuit technology has been largely underused in the context of neuroscience studies mainly because of the inherent difficulty in implementing and tuning this type of interactions. In this paper, we present a set of algorithms for the automatic adaptation of model neurons and connections in the creation of hybrid circuits with living neural networks. The algorithms perform model time and amplitude scaling, real-time drift adaptation, goal-driven synaptic and model tuning/calibration and also automatic parameter mapping. These algorithms have been implemented in RTHybrid, an open-source library that works with hard real-time constraints. We provide validation examples by building hybrid circuits in a central pattern generator. The results of the validation experiments show that the proposed dynamic adaptation facilitates closed-loop communication among living and artificial model neurons and connections, and contributes to characterize system dynamics, achieve control, automate experimental protocols and extend the lifespan of the preparations