Information theoretic analysis of brain signals

One of the fundamental challenges of Neuroscience is to understand how the brain processes, codes and communicate information using the concerted activity of large ensembles of nerve cells. Thanks to recent advances in technology it is now possible to record this population activity directly in animals and indirectly in humans. However, there is still little knowledge on how to analyze these signals to understand what computation takes place inside the brain. The proposed mathematical research aims at generating new advanced mathematical algorithms and computer programs that can be applied to recordings from the brains of animals and humans to decipher the messages that are hidden into the electrical signals generated by the brain. The application of these new mathematical techniques will increase our understanding of how healthy brains work, a fundamental basic step toward curing diseases of the nervous system. The ability of interpreting and deciphering the meaning of electrical signals taken from the brain will also be important to develop neural prostheses that could restore motor or sensory functions in patient with brain or spinal cord damages.

The goal of the proposed research is to understand how sensory information is encoded in the activity of large neuronal populations. Neuronal population activity can be accessed directly as action potentials (or spikes) simultaneously recorded from several cells, or indirectly as field potential signals. These include Local Field Potentials (LFPs) or Magneto-Encephalograms (MEGs), and reflect the average electric field generated by whole neuronal ensembles. In cortex, they are likely to reflect mostly synaptic inputs to the area and local cortical processing.
The general Neuroscientific aim of this project is to understand how these different types of electrical neural activity encode information, and will be met by working on the following specific areas. First, I will develop robust data analysis methods based on Information Theory, which will allow the discovery of how sensory information is encoded in both spikes and field potentials. Second, I will apply these new methods to visual and somatosensory recordings from our collaborators Logothetis, Rainer, Diamond, Petersen and Green, to provide the first rigorous characterization of how large neuronal populations encode information. Third, I will use neural network models of cortical areas to characterize the biophysical mechanisms and the functional anatomy underlying the generation of both field potentials and spikes. This model will be used to study in detail under which conditions spikes and field potentials convey either the same or complementary information, and to characterize what aspects of functional neuronal interactions can be extracted from the simultaneous analysis of these different signals. Taken together, the mathematical methods, the data analysis and the modelling developed here will significantly enlarge the impact on Neuroinformatics and systems-level neuroscience research on neural codes.