Maximum entropy model estimates functional connectivity

Tools to estimate brain connectivity have been useful to improve our
understanding of brain functioning. Reduced models ofcultured neurons are
often used to study the behavior of neuronal networks, including
functional connectivity and how it mightbe affected by external stimuli.
Cultured neurons tend to be active in ensembles, and when pairs of neurons
show significantsynchronicity in their firing patterns they are said to be
functionally connected. The most common methods to infer
functionalconnections are based on pair-wise cross correlation between
activity patterns of (small groups of) neurons. However, thesemethods are
not designed to be used during external stimulation, and they are
relatively insensitive to inhibitory connections.Maximum Entropy (MaxEnt)
models may provide a conceptually different method to infer functional
connectivity, with thepotential benefit to estimate functional
connectivity in the presence of an external stimulus and to infer
excitatory as well asinhibitory connections. These models do not use
pairwise comparison, but are based on probability distributions of sets
ofneurons that are synchronously active in discrete time bins. We
investigate the ability of the MaxEnt models to infer
functionalconnectivity, using electrophysiological recordings fromin
vitroneuronal cultures on micro electrode arrays. We comparefunctional
connectivity as inferred by MaxEnt models to that obtained by conditional
firing probabilities (CFP), an establishedcross-correlation based method.
We show that MaxEnt models provide connectivity estimates that correlate
well with CFPoutcomes. In addition, stimulus-induced connectivity changes
were detected by MaxEnt models, and were of the samemagnitude as those
detected by CFP.