Detection of Stimuli Changes in Neural Eventograms Using the Line of Synchronization of Global Recurrence Plots

Reliable detection of stimulus-driven states and their separation from internal state-driven spontaneous activity is an important step towards inferring temporal dynamics of neurons and its relation to the perception of external inputs. This is challenging, especially when no prior assumptions about the underlying model and data generating processes exist. To address this task, we applied efficient recurrence quantification analysis (RQA) based on global recurrence plots (RP) for accurate identification of the onset and offset of visual neuronal responses caused by distinct types of visual stimuli. In particular, these critical times are estimated by taking the first order difference of the line of synchronization extracted from the associated global RP. Our approach was evaluated using a real dataset of visually-driven neuronal responses and spontaneous activity (recorded by in vivo 2-photon calcium imaging). It accurately detects both the onset and offset time instants in the eventograms of pyramidal neurons in a completely model agnostic framework.