Prednáška Dr. Gabriely Czanner

Dňa 28.4. 2008 o 14:00 sa v zasadačke ústavu uskutoční seminár Dr. Gabriely Czanner z Warwick Medical School and Warwick Manufacturing Group, University of Warwick, Coventry, United Kindgom na tému "Point Process Statistical Framework for Improved Inference from Neural Spike Train Observations".

Abstract:

Understanding how neurons respond to stimuli is a fundamental question in neuroscience. It is believed that neurons communicate by encoding information about a stimulus via short electrical discharges, called spikes. The duration of each spike is very short, on the order of microseconds, and experimenters generally regard these events as binary, i.e. at a particular point in time, an action potentially either did or did not occur. The collection times of spikes is termed a spike train. A crucial property of neuronal responses (i.e. the spike trains) is that if we apply same stimulus twice, the responses are similar but not the same due to neurons being inherently noisy. Therefore a natural way to study spike trains is in the framework of point stochastic processes, a framework specifically designed for the analysis of binary stochastic events. Yet neuroscientists mostly rely on histogram-based ANOVA methods to analyze the spike train data.

We present a state-space generalized linear model (SS-GLM) to formulate a point process representation of between-trial and within-trial neural spiking dynamics in multiple trials. To estimate the model parameters by approximate Expectation-Maximization algorithm we use a recursive point process filter and fixed-interval smoothing algorithm as analogs of Bayes’ filter and smoother; and we implement statistical inference to answer neurophysiological questions. We illustrate our approach in two applications. In the analysis of hippocampal neural activity recorded from a monkey, we use the model to quantify the neural changes related to learning. In the analysis of primary auditory cortical responses to different levels of electrical stimulation in the guinea pig midbrain, we propose more sensitive auditory threshold detection. Our findings have important implications for developing theoretically-sound and practical tools to characterize the dynamics of spiking activity.

The presented work has been done while the presenter was a postdoctoral research fellow in the Neuroscience Statistics Research Lab of Prof Emery Brown in Massachusetts General Hospital and Harvard Medical School. This work has been done in collaboration with Anna A Dreyer from Massachusetts Institute of Technology, USA; Uri T Eden from Brown University, USA; Sylvia Wirth from Centre National de la Recherche Scientifique, France; Hubert H Lim from Hannover Medical School, Germany; Marianna Yanike and Wendy Suzuki from Center for Neural Science at New York University.