Robust analysis of non-stationary cortical responses: Tracing variable frequency gamma oscillations and separating multiple component input modulations

Abstract

Natural signals have a complex nature and most often encode information both in the frequency and time domain. Neuronal signals in particular have a very nonlinear behavior, with features of interest appearing sparsely and discontinuously. Therefore, methods that characterize and enable the visualization of these data are of great importance. Here, we present two algorithms that act on different dissociation problems in neuroscience: Firstly, the definition of trajectories in a time-frequency-power three dimensional space and secondly, the dissociation of modulatory effects of different time scales. The methods have the advantage of preserving the transient nature of neuronal features and of providing a practical computational implementation. We apply these methods on cortical response to visual stimuli where multiple parameters are manipulated. Results show that such response characteristics reveal features that are not explained by current theories of underlying mechanisms of oscillatory response.