Experimental – Modelling approach to study communication efficiency in neurohybrid systems

At the intersection of neuroscience, bioengineering, and information technol- ogy, neurohybrid systems represent a rapidly expanding field. By enabling bidirectional communication between biological networks and artificial de- vices, these systems hold promise not only for therapeutic applications such as intelligent neuroprosthetics, cochlear and retinal implants, or adaptive deep brain stimulation, but also for advancing fundamental brain research, neuro- morphic computing, and biocomputing. Despite this potential, a key challenge remains: there is no standardized framework for efficient neural stimulation. Current approaches often rely on empirical rules, while most models of the neuron–device interface simplify critical aspects such as adhesion, cleft mor- phology, and extracellular electric fields. This thesis addresses these gaps by combining modeling and experimental approaches to investigate the mecha- nisms governing communication efficiency in in vitro neurohybrid systems. We explored four different scenarios, each exploring typical neurohybrid configurations from a different point of view. First, we studied capacitive elec- troporation, showing that cell–substrate geometry and cleft resistance trans- form the system into a band-pass filter, making signal choice case-specific. We demonstrated that electroporation efficiency depends on the time spent within a specific transmembrane voltage range rather than on absorbed energy, high- lighting the importance of protocol design. This result emphasizes the im- portance of waveform design and suggests that pre-experiment modeling can greatly improve reproducibility and reduce experimental trial-and-error. Sec- ond, we developed a model of the neuromuscular junction, emphasizing the role of Schwann cells and ephaptic feedback in enhancing transmission reli- ability. Our results suggest that extracellular resistance and electric insula- tion are key mechanisms that were evolved to ensure robustness even under pathological conditions. These findings point at Schwann cells as possible ther- apeutic interventions targets to slow down the progression of degenerative diseases. Third, we proposed and experimentally demonstrated, for the first time, that lithium niobate can be used as a novel neurostimulation platform. By leveraging its photovoltaic properties, we showed that both excitatory and inhibitory responses can be achieved depending on cell geometry and laser po- sitioning. Modeling also guided experimental design, enabling the definition of efficient protocols, which we employed to acquire a large dataset of elec- trophysiological recordings. This work proves the biocompatibility of lithium niobate and its possible employment in electrophysiological experiments for the first time. Finally, we introduced an energy-based analytical framework for next-generation neural mass models. By treating these systems as perturbed Hamiltonian systems, we linked potential energy landscapes to neural activ- ity, offering a novel and physiologically grounded interpretation of population responses to external inputs. This work is a first step toward bridging the gap between energy landscape theory and physiological network models. Taken together, these studies reveal common principles underlying neu- ral communication across biological and neurohybrid contexts. In particular, they highlight the central role of the cell–substrate cleft, extracellular fields, and geometry-dependent filtering in shaping effective interactions. Beyond mechanistic insights, this work demonstrates how modeling can act as a unify- ing framework to interpret results, optimize stimulation protocols, and guide experimental design. The ability to anticipate outcomes reduces costs, accel- erates discovery, and minimizes reliance on animal-derived material, thereby contributing to more ethical and sustainable research practices.