Neural Information Encoding

Spatial Information in the Rat Hippocampus

   This project focuses on how individual and ensembles of neurons encode information about relevant biological stimuli, and uses point process models of spatial information representation by neurons in the rat hippocampus. This work is in collaboration  with Dr. Matthew Wilson at MIT, a pioneer in obtaining simultaneous multiple single spike train recordings of pyramidal (place) neurons cells in the hippocampus of freely-behaving rodents, and with Dr. Loren Frank at UCSF, who is carrying out similar experiments in his lab.

    The primary specific aim is to study how pyramidal cells in the rat hippocampus maintain and continually update a representation of the animal’s position in the environment. The benefits of this research are: a) a real-time assessment of information processing in the rat hippocampus; b) an improved quantitative understanding of one of the mechanisms by which information is encoded into short-term memory; c) development of a general paradigm for analyzing how information is encoded in neural systems data.

    The research has initially focused on modeling the single cell by using non-Poisson stimulus-response models of neural spike train activity. We used our models to estimate the dynamics of both the spatial receptive field (spatial intensity function) and the interspike interval structure (temporal intensity function) of neural spike trains on a millisecond time scale. We applied the Bayesian algorithms on CA1 place cells recordings from rats foraging in an open environment. Results demonstrated that these models provide more accurate descriptions of these local spiking properties of the neuron.  This initial characterization has been published in the Journal of Neuroscience Methods (Barbieri et al. 2001) and Neurocomputing (Barbieri et al. 2002).

     In light of these new findings, we were able to refine the Bayesian decoding algorithms introduced in Brown et al (1998), for position estimation of the same open environment recordings considered in Barbieri at al. 2001 and 2002. Using only the recorded activity of 30 cells, our most refined paradigm is now able to obtain position estimation with errors as small as 5.5 cm from the actual values (from the previous 6.7 cm), with an average coverage probability of 0.75 (from the previous 0.34). We suggest our point process linear state-space model framework as an approach to dynamic signal processing for neural systems. This work appeared in Neural Computation (Barbieri et al. 2004). We have also focused attention to models able to consider more complex history dependence, multivariate interactions among cells, as well as temporal changes in neuronal activity in response to adaptation and learning. To this extent, we have applied multivariate analysis for calculation of synaptic coupling coefficients. Preliminary results allow for the first attempt in constructing a neuronal network from single cell recordings, which also takes into account the influence from Theta Rhythm modulation. We have further explored generalizations of adaptive filter algorithms based on maximum a posteriori and sequential Monte Carlo filters (Ergun et al. 2007) that can be used for both adaptive estimation of neural plasticity and ensemble neural spike train decoding.  Such algorithms have been also considered in our lab for other neural data. 

More recently, research has focused on characterization of up-down states in neurons from the primary somatosensory cortex in behaving rats, and of fast oscillations (ripples) in the rodent hippocampus.

Selected Publications

Barbieri R, Quirk MC, Frank LM, Wilson MA, Brown EN. A time-dependent analysis of spatial information encoding in the rate hippocampus. Neurocomputing, 2000, 32-33: 629-635.

Barbieri R, Quirk MC, Frank LM, Wilson MA, Brown EN. Construction and analysis of non-Poisson stimulus response models of  neural spike train activity. J Neurosci. Methods, 105 (1):25-37, 2001.

Barbieri R, Quirk MC, Frank LM, Wilson MA, Brown EN. Diagnostic methods for statistical models of place cell spiking activity. Neurocomputing, 38:1087-1093, 2001.

Barbieri R, Frank LM, Quirk MC, Wilson MA, Brown EN. Construction and analysis of non-Gaussian place field models of neural spiking activity. Neurocomputing, 44-46: 309-314, 2002.

Eden UT, Frank LM, Barbieri R, Solo V, Brown EN. Adaptive filtering algorithms for neural spike trains. Neurocomputing, 44-46: 309-314, 2002.

Brown EN, Barbieri R, Ventura V, Kass RE, Frank LM. The time-rescaling theorem and its application to neural spike train data analysis. Neural Computation, 14: 325-346, 2002.

Brown EN, Barbieri R, Eden UT, Frank LM. Likelihood methods for neural data analysis. In: Feng J, ed. Computational Neuroscience: A Comprehensive Approach. London: CRC, pp. 253-286, 2003.

Barbieri R, Frank LM, Nguyen DP, Quirk MC, Solo V, Wilson MA, Brown EN. Dynamic analyses of information encoding in neural ensembles. Neural Computation, 16: 277-307, 2004.

Eden UT, Frank LM, Barbieri R, Solo V, Brown EN. Dynamic analysis of neural encoding by point process adaptive filtering. Neural Computation, 16: 971-998, 2004.

Barbieri R, Frank LM, Wilson MA, Brown EN. An analysis of hippocampal spatio-temporal representations using a Bayesian algorithm for neural spike train decoding. IEEE Transactions on Neural Systems and Rehabilitation Engineering, Special Issue on Neural Engineering, 13: 131-136, 2005.

Brown EN and Barbieri R. Dynamic analyses of neural representations using the state-space modeling paradigm. In: Madras B, Von Zastrow M, Colvis C, Rutter J, Shurtleff D, Pollock J. The Cell Biology of Addiction, New York: Cold Spring Harbor Laboratory Press, pp. 415-432, 2005.

Ergun A, Barbieri R, Eden UT, Wilson MA, Brown EN. Construction of point process adaptive filter algorithms for neural systems using sequential Monte Carlo methods. IEEE Transactions on Biomedical Engineering, 54(3): 419-428, 2007.