Information Processing in Neuromorphic Biohybrids
Information Processing in Neuromorphic Biohybrids
Where did the research take place?
The study site has not been established. Author addresses may differ from where the research occurred.
A plain-language reading has not been prepared for this paper yet.
Original abstract
The field of Brain-Machine interfaces (BMI), i.e. connecting animal or human brains to some technological processing system, has seen increasing popularity in recent years. Usually, these are targeted at specific applications, such as deep brain stimulation for Parkinson treatment, or motor prosthesis replacements of a missing limb. Processing at the technological side is aimed at solving a specific task, not necessarily at understanding better the interface or developing a more general theory of brain-machine interaction. However, there is also a growing community that aims to develop so-called biohybrids, where the technological side acts like an extension or real counterpart of biological tissue, being more tightly coupled in a closed loop with the tissue instead of the usual open loop approaches in BMI. This is aided by significant advances in recent years in both the theoretical aspects of understanding the dynamics of spiking neural networks and a progress in electrophysiological methods that is pushing the frontiers of intelligent neural interfacing and signal processing technologies. The second ingredient driving biohybrids is recent advance in neuromorphic circuits. Neuromorphic circuit design is a branch of semiconductor circuit design that implements circuits which mimic aspects of the behaviour of living nerve tissue. Popular aspects to be implemented are e.g. pulsing communication between neurons and biological behaviour in neurons such as regular spiking, accommodating, bursting, etc. Plasticity is also implemented often, e.g. in neurons as spike frequency adaptation or in synapses as facilitation/depression or as longer-term learning such as spike-time-dependent plasticity. Thus, neuromorphic circuits are a natural choice for an interface with in-vitro or in-vivo neural tissue, providing a loop back from biology-like circuit design to actual biology. Bluring the boundaries between the technological and the biological side in biohybrid systems is projected to lead to a significantly deeper understanding of information representation and -processing in neurobiology, providing an exciting experimental framework for engineers and neurophysiologist and thus ultimately leading to completely new classes of neuroprostheses and brain-machine interfaces. This thesis details the contribution of the author to two different biohybrids in two European projects, one more targeted at the single neuron level (Ramp project), the other targeted at the population level (Coronet project). Circuit design, system integration work and measurements for both are given. It is demonstrated that especially the integration of certain types of dynamics and plasticity at the neuron as well as population level lead to an improved coupling of the technological and biological side of the biohybrid, with information transferred in both directions enriched by being able to couple into the respective dynamics instead of doing a static decoding as in conventional BMI. For instance, for the population-level biohybrid in the Coronet project, the transfer of detailed network dynamics such as population up and down states and their substructure is demonstrated, which matches well with theories of the brain operating in a noisy attractor framework. $\textit{Full-text embargoed until: 2025-09-06}$