Application of Poisson-based hidden Markov models to in-vitro neuronal data

Recent advances in electrophysiological techniques have made it possible to culture in vitro biological networks and closely monitor ensemble neuronal activity using multi-electrode recording techniques. One of the main challenges in this area of research is attempting to understand how intrinsic activity is propagated within these neuronal networks and how it may be manipulated via external stimuli in order to harness their computational capacity. This raises the question of what similarities and differences arise between spontaneous and evoked responses and how external stimulation can be optimally applied in order to robustly control the neuronal plasticity of neuronal cultures. In this paper we present in detail an application of machine learning methods, specifically hidden Markov models with Poisson-based output distributions, with which we aim to perform comparative studies between spontaneous and evoked neuronal activity over different ages of network development.