Detection of bursts in extracellular spike trains using hidden semi-Markov point process models

Neurons in vitro and in vivo have epochs of bursting or “up state” activity during which firing rates are dramatically elevated. Various methods of detecting bursts in extracellular spike trains have appeared in the literature, the most widely used apparently being Poisson Surprise (PS). A natural description of the phenomenon assumes (1) there are two hidden states, which we label “burst” and “non-burst,” (2) the neuron evolves stochastically, switching at random between these two states, and (3) within each state the spike train follows a time-homogeneous point process. If in (2) the transitions from non-burst to burst and burst to non-burst states are memoryless, this becomes a hidden Markov model (HMM). For HMMs, the state transitions follow exponential distributions, and are highly irregular. Because observed bursting may in some cases be fairly regular—exhibiting inter-burst intervals with small variation—we relaxed this assumption. When more general probability distributions are used to describe the state transitions the two-state point process model becomes a hidden semi-Markov model (HSMM). We developed an efficient Bayesian computational scheme to fit HSMMs to spike train data. Numerical simulations indicate the method can perform well, sometimes yielding very different results than those based on PS.

[1]  Jun S. Liu,et al.  The Multiple-Try Method and Local Optimization in Metropolis Sampling , 2000 .

[2]  C. Legéndy,et al.  Bursts and recurrences of bursts in the spike trains of spontaneously active striate cortex neurons. , 1985, Journal of neurophysiology.

[3]  Satish Iyengar,et al.  Modeling neural activity using the generalized inverse Gaussian distribution , 1997, Biological Cybernetics.

[4]  Christof Koch,et al.  Biophysics of Computation: Information Processing in Single Neurons (Computational Neuroscience Series) , 1998 .

[5]  Michael W. Levine,et al.  The distribution of the intervals between neural impulses in the maintained discharges of retinal ganglion cells , 1991, Biological Cybernetics.

[6]  Faming Liang,et al.  A Theory for Dynamic Weighting in Monte Carlo Computation , 2001 .

[7]  Brent Doiron,et al.  Inhibitory feedback required for network oscillatory responses to communication but not prey stimuli , 2003, Nature.

[8]  J. Cowan,et al.  Excitatory and inhibitory interactions in localized populations of model neurons. , 1972, Biophysical journal.

[9]  John M. Olin.  Calculating posterior distributions and modal estimates in Markov mixture models , 1996 .

[10]  R. Dykes,et al.  Recognition of temporally structured activity in spontaneously discharging neurons in the somatosensory cortex in waking cats , 1997, Brain Research.

[11]  Vivien A. Casagrande,et al.  Biophysics of Computation: Information Processing in Single Neurons , 1999 .

[12]  David C. Tam.  An alternate burst analysis for detecting intra-burst firings based on inter-burst periods , 2002, Neurocomputing.

[13]  L. Baum,et al.  Statistical Inference for Probabilistic Functions of Finite State Markov Chains , 1966 .

[14]  Matthew A. Wilson,et al.  Discrete- and Continuous-Time Probabilistic Models and Algorithms for Inferring Neuronal UP and DOWN States , 2009, Neural Computation.

[15]  M. Corner,et al.  Physiological effects of sustained blockade of excitatory synaptic transmission on spontaneously active developing neuronal networks—an inquiry into the reciprocal linkage between intrinsic biorhythms and neuroplasticity in early ontogeny , 2002, Neuroscience & Biobehavioral Reviews.

[16]  Naftali Tishby,et al.  Cortical activity flips among quasi-stationary states. , 1995, Proceedings of the National Academy of Sciences of the United States of America.

[17]  Emery N. Brown,et al.  The Time-Rescaling Theorem and Its Application to Neural Spike Train Data Analysis , 2002, Neural Computation.

[18]  Boris Gourévitch,et al.  A nonparametric approach for detection of bursts in spike trains , 2007, Journal of Neuroscience Methods.

[19]  Lawrence R. Rabiner,et al.  A tutorial on hidden Markov models and selected applications in speech recognition , 1989, Proc. IEEE.

[20]  Frank C. Hoppensteadt,et al.  Bursts as a unit of neural information: selective communication via resonance , 2003, Trends in Neurosciences.

[21]  Jianfeng Feng,et al.  Computational neuroscience , 1986, Behavioral and Brain Sciences.

[22]  S. L. Scott.  Bayesian Analysis of a Two-State Markov Modulated Poisson Process , 1999 .

[23]  Emery N. Brown,et al.  Computational Neuroscience: A Comprehensive Approach , 2022 .

[24]  Riccardo Barbieri,et al.  Probabilistic models and inference algorithms for neuronal decoding of UP and DOWN states , 2008, BMC Neuroscience.

[25]  N. Spruston,et al.  Output-Mode Transitions Are Controlled by Prolonged Inactivation of Sodium Channels in Pyramidal Neurons of Subiculum , 2005, PLoS biology.

[26]  Lon Turnbull,et al.  The string method of burst identification in neuronal spike trains , 2005, Journal of Neuroscience Methods.

[27]  S. Sherman,et al.  Intracellular and extracellular in vivo recording of different response modes for relay cells of the cat's lateral geniculate nucleus , 2004, Experimental Brain Research.