The ability to predict the onset of epileptic activity remains an important goal of preclinical neuroscience. Here, we investigate the transition from spontaneous Up/Down states, considered the default cortex network activity, to the hypersynchronous epileptiform state. We recorded local field potential (LFP) activity in acute cortical slices of wild-type and transgenic mouse models and used various seizure-induction protocols. The PSR method, a powerful tool in the field of nonlinear dynamical systems, has been used to gain insights of this dynamic complex phenomenon. It has been shown that starting even from one-dimensional time series, we can create a Reconstructed Phase Space which preserves the dynamical properties of the underlying system generating the signal. For the analysis of the reconstructed trajectories, we have used several suitable metrics such as Lyapunov exponents, fractal dimensionality metrics, several entropies and informational complexity quantifiers. Our results reveal the capacity to discriminate between spontaneous, pre-ictal and seizure-like states, for several metrics and independently of the activity level of the signal. They can also discriminate between spontaneous and pre-ictal states when only the events or only the quiescent segments of the trajectories are considered. Collectively, the observed differences indicate that as the system is heading towards epilepsy, its behaviour gradually becomes less complex, and suggest that the relevant entropic changes calculated through PSR could be used as a means of predicting seizure-like states.