Computational neuroscience is an interdisciplinary discipline in which modelling and analysis tools derived from mathematics, physics and computer science are used to investigate how the nervous system processes information. It mostly relies on the development, simulation, and analysis of multi-scale models of brain function, from the level of molecules through single neurons and neuronal networks up to cognition and behaviour. The analysis of real electrophysiological data recorded from various locations of the brain on different temporal and spatial scales is used to validate the models.

Our computational neuroscience work in Florence goes along two main lines of research, theoretical analysis of neuronal networks and data analysis. The first line of research relies mostly on the modelling and simulation of neuronal networks using simple models (such as Integrate and Fire and Hodgkin-Huxley for the single neurons or Wilson-Cowan at the population level). Currently, particular attention is devoted to research on hub cells, which are neurons able to strongly impact and control the network dynamics. The second line of research deals with the analysis of electrophysiological recordings such as EEG and neuronal spike trains. Here a recent focus of interest is neuronal population coding, i.e., the study of how the sensory world is represented in the action potentials of neuronal networks in the brain.

Recent publications:

Olmi S, Petkoski S, Guye M, Bartolomei F, Jirsa V:

PLoS computational biology 15(2): e1006805 (2019)

Epilepsy is characterized by perturbed dynamics that originate in a local network before spreading to other brain regions. In this paper we studied studied patient-specific brain network models of epilepsy patients, comprising 88 nodes equipped with region specific neural mass models capable of demonstrating epileptiform discharges. Applying stability analysis led to a seizure control strategy that is significantly less invasive than the traditional surgery, which typically resects the epileptogenic regions. The invasiveness of the procedure correlates with graph theoretical importance of the nodes. The novel method subsequently removes the most unstable links, a procedure possible by advent of novel surgery techniques. Our approach is entirely based on structural data, allowing creation of a brain model based on purely non-invasive data prior to any surgery.

Satuvuori E, Mulansky M, Daffertshofer A, Kreuz T:

JNeurosci Methods 308, 354 (2018) and arXiv [PDF]

This article simulates how neuronal populations in the brain work together to distinguish different sensory inputs from the real world (e.g. visual images). More specifically, it proposes two new algorithms (one where each neuron acts on its own and one where they all work together) to find among a large neuronal population the one subpopulation that discriminates different stimuli best.

S. Luccioli, D. Angulo-Garcia, R. Cossart, A. Malvache, L. Módol-Vidal, V. H. Sousa, P. Bonifazi, A. Torcini:

PLoS Comput. Biol. 14(11) 2018 [PDF]

In this article we studied the mechanisms underlying the presence of synchronous activity in developing neural circuits. In particular, we developed a neural network model which reproduces recent experimental findings in the neo-cortex, i.e. the existence of peculiar neurons (called "drivers") which, under stimulation, have the capability to change the frequency of the synchronization events of the overall neural population.