Idan Segev

Title: Towards an Objective Analysis of the Firing Variability of Cortical Neurons

Abstract: The existence of a diversity of electrical classes of cortical neurons is widely agreed upon yet is primarilly based on subjective categorization. Classes such as “regular spiking”, “fast-spiking”, “stutterers”, “bursters” were identified based on the response of neurons to application of depolarizing current steps. Here, analyzing a large database of spiking responses of hundreds of cortical neurons to a standardized set of stimuli, including current pulses, ramp pulses and noisy inputs, we developed methods to objectively assess whether indeed different electrical classes could be identified. These methods include combinations of different dimensionality reduction algorithms (e.g. principle component analysis), machine learning applications and information theoretic tools (e.g. mutual information between different variables). Together these methods are used for quantitatively characterizing the firing pattern of a neuron based on “features” of the spiking discharge, such as spike width, time-to-first-spike, spike frequency, degree of “burstiness”, etc. Extracting these features from the database of cells enables us to take into account the variability of these features across different stimuli, different neurons and different putative electrical classes, thus, allowing us to investigate the existence of electrical classes and their coherence in the face of different stimuli. We initially show that parts of the accepted subjective classification can be recovered using observer independent analysis. However, the classification clearly depends on the features used as the basis of the classification. Thus, we then proceed to consider the utility of different features related to different stimuli through statistical and information theoretic tools. Ultimately, we aim to suggest a “most informative stimulus” set, along with the corresponding features.

Bio sketch: Idan Segev's research team utilizes computational tools ranging from cable theory to compartmental modeling to statistical methods and information theory to study how neurons, the elementary microchips of the brain, compute and dynamically adapt to our ever-changing environment. More recently, he has worked jointly with several experimental groups worldwide in an endeavor to model in detail the cortical column – a functional unit containing thousands of intensely but very specifically connected networks of neurons. This project also aims at developing automated methods for generating models of the different electrical and morphological classes of neurons found in the column. The ultimate goal is to unravel how local fine variations within the cortical network underlie specific computations (e.g., the orientation of a bar in the visual system) and may give rise to certain brain diseases or to a healthy (and “individual”) brain. Idan Segev is the David & Inez Myers Professor in Computational Neuroscience and former director of the Interdisciplinary Center for Neural Computation (ICNC) at the Hebrew University of Jerusalem.