Emery Brown


First Name: Emery
Last Name: Brown


Mathematical biologist Emery N. Brown and neuroscientist Steven Massaquoi are collaborating in a joint venture, The Neuroscientist Advisory Board, see http://www.tnsab.com/Company/Advisors/advisors.html.


Professor of Computational Neuroscience and Health Sciences

and Technology
Department of Brain and Cognitive Sciences
MIT-Harvard Division of Health Science and Technology
Massachusetts Institute of Technology
77 Massachusetts Avenue, 46-6079
Cambridge, MA 02139
tel: 617 324 1879

Massachusetts General Hospital Professor of Anaesthesia
Harvard Medical School
Department of Anesthesia and Critical Care
Massachusetts General Hospital
55 Fruit Street Clinics 3
Boston, MA 02114
tel: 617 726 7487
fax: 617 726 8410
email: brown@neurostat.mgh.harvard.edu


Neural Information Encoding

The study of how neural systems encode information is a fascinating and challenging problem. An active area of research in neuroscience is the development of encoding and decoding algorithms: mathematical techniques to decipher how ensembles of neurons represent and transmit information about pertinent signals from the outside world. The point process character of neural spike trains means that standard signal processing techniques developed to analyze continuous signals will have limited application in the analysis of neural systems. The study of neural signal processing requires the development of new quantitative techniques to accurately characterize the properties of neural systems. These methods are being developed in collaboration with Professor Matthew A. Wilson in the Department of Brain and Cognitive Science at MIT. The research uses Professor Wilson's paradigm of simultaneous multiunit recordings of pyramidal (place) cells in the hippocampus of freely behaving rodents to model how these neurons encode spatial information as the animals perform specific behavioral tasks. Located on the inner surface of the brain's temporal lobes, the hippocampus is a region critical for the formation and storage of both short- and long-term memories. It is now appreciated that within the hippocampus the rat uses specialized neurons known as place cells to develop a spatial map of an environment. As an animal moves through a new environment, each hippocampal place cells demarcates, within as few as 5 minutes, its own region in the environment by firing spikes only when the animal is within that region. The region of the environment in which the cell fires is termed its place field. Large numbers of hippocampal place cells tile each environment with overlapping place fields and their ensemble firing patterns gives a continuous representation of the animal's spatial location. Spiking activity of the hippocampal place cells is also modulated by the 6 to 14 Hz theta rhythm. The objective of the research is to develop accurate statistical models of individual and ensemble place cell spiking activity. These models are being used to study the dynamics of individual and ensemble place cell representations of spatial information under various behavioral and environmental paradigms. The statistical methods we are developing are based on spatio-temporal point process models of hippocampal place cell spiking activity, parametric maximum likelihood methods, point process adaptive filtering techniques and nonlinear recursive filtering algorithms based on Bayesian statistical theory.

Adaptive Point Process Models of Receptive Field Plasticity

In addition, we have also begun developing adaptive point process models of spike train activity that seek to track plasticity related changes in neural receptive fields. The algorithm updates its representation of the receptive field on a millisecond timescale and is able to accurately track the dynamics of a simulated place cell whose field changes over time. The two links below show animations of the tracking of a simulated and a real CA1 place cell using a Gaussian place field and an assumption of Poisson spike train statistics. In the simulated example, the rat is represented by the yellow dot, the yellow curve shows the actual place field shape, and the green curve shows the algorithms instantaeous estimate of the place field. The red lines represent spikes and the number of the left side of the animation shows the time in seconds. In the real example, only the estimated field is shown. In both cases, the algorithm uses the first 50 spikes to obtain a maximum likelihood estimate of the place field parameters and the estimation begins after the 50th spike. We are currently working on models with more flexible spatial and temporal components that allow us to more accurately model both the shape of the place field and the temporal dependencies found among adjacent spikes (i.e. theta rhythmicity and bursting).

Statistical Modeling of Functional Neural Imaging Data

Functional magnetic resonance imaging (fMRI) offers a means to visualize directly changes in cerebral physiology with highly accurate spatial and temporal resolution. Changes in the fMRI signal are related to changes in cerebral blood flow, blood volume, oxygen consumption and level of neuronal activity. Similiarly, optical imaging is a rapidly developing technique for making dynamic images of cerebral cortical function. One of the greatest strengths of fMRI is the ability to design and link data acquisition and analysis strategies. The long-term objective of this research is to improve the flexibility and rigor of fMRI data analysis strategies using statistically motivated presentation paradigms and data analysis techniques. The goals of the project are to: (1) develop methods for event-related fMRI with rapid, randomized stimulus presentation designs; (2) develop a physiologically-based statistical model and estimation procedure for the signal and noise in fMRI, and (3) test the utility of these tools for better optimizing and quantifying fMRI experiments. This research is one of the four foundation projects that make up the recently established Center for Functional Neuroimaging Technologies (CFNT), headed by Dr. Bruce Rosen at the MGH NMR Center. CFNT is funded by the National Center for Research Resources. The other three projects are headed by Dr. John Belliveau, Dr. David Boas and Dr. Anders Dale.

 Statistical Modeling of Circadian and Neuroendocrine Rhythms

Circadian rhythms are biological rhythms generated by an organism or group of organisms that have an intrisic period of approximately 24 h. The term circadian was coined by Franz Halberg from the Latin circa, meaning about and dies meaning day. In human the site of the human circadian pacemaker or biological clock is the suprachiasmatic nucleus of the hypothalamus. This approximately 24 h oscillator helps ensure correct timing among the body's physiological processes and between those processes and events in the outside world. The human circadian pacemaker is studied by measuring marker rhythms whose behaviors are known to be tightly coupled to the clock.The three principal marker rhythms used in human circadian studies are core temperature, plasma cortisol levels, and plasma melatonin levels. This research is a collaboration with Dr. Gail Adler and Dr. Charles A. Czeisler in the Circadian, Neuroendocrine and Sleep Disorders Section of the Division of Endocrinology in the Department of Medicine at Brigham and Women's Hospital. Some of the questions we are investigating are: computing accurate estimates of the period, phases and amplitude of the human circadian pacemaker from different marker rhythms (view article); accurately decomposing neuroendocrine rhythms into their circadian, ultradian, and kinetic components; quantifying the effects of lighton the circadian pacemaker; and characterizing differences between normal physiology and disease states in terms of the components of neuroendocrine rhythms. The models for core-temperature are based on modified versions of the weakly nonlinear van der Pol differential equation and a continuous time autoregressive model. The models for the neuroendocrine rhythms are both stochastic and deterministic linear differential equations with pulse inputs.

Selected Publications

Neural Information Encoding

  1. Brown EN, The theory of point processes for neural systems, Lecture Notes in Theoretical Physics, 2003 Summer School, Les Houches, France, 2004, In Press.
  2. Smith AC, Frank LM, Wirth S, Yanike M, Hu D, Kubota Y, Graybiel AM, Suzuki W, Brown EN. Dynamic analysis of learning in behavioral experiments, Journal ofNeuroscience, 2004, 15:965-91.
  3. Barbieri R, Frank LM, Nguyen DP, Quirk MC, Solo V, Wilson MA, Brown EN, Dynamic analyses of information encoding by neural ensembles. Neural Computation, 2004, 16 (2): 277-307.
  4. Eden UT, Frank LM, Barbieri R, Solo V, Brown EN, Dynamic analyses of neural encoding by point process adaptive filtering, Neural Computation, 2004, 16(5), 971-998.
  5. Brown EN, Kass RE, Mitra PP, Multiple neural spike train data analysis: state-of-the-art and future challenges, Nature Neuroscience, 2004, 7(5): 456-61.
  6. Smith AC, Brown EN. Estimating a state-space model from point process observations. Neural Computation. 2003;15:965-91.
  7. Nguyen D, Frank LM, Brown EN. An application of reversible-jump MCMC to spike classification of multiunit extracellular recordings. Network: Computation in Neural Systems 2003;14:61-82..
  8. Brown EN, Barbieri R, Eden UT, Frank LM. Likelihood methods for neural data analysis. In Press. In: Feng J, ed. Computational Neuroscience: A Comprehensive Approach, London: CRC, 2003; p. 253-86.
  9. Wirth S, Yanike M. Frank LM, Smith AC, Brown EN, Suzuki WA. Single neurons in the monkey hippocampus and learning of new associations. Science 2003. 300;1578-81.
  10. Brown EN, Barbieri R, Ventura V, Kass RE, Frank LM. The time-rescaling theorem and its application to neural spike data analysis. Neural Computation 2002. 14(2): 325-46
  11. Frank LM, Eden UT, Solo V, Wilson MA, Brown EN. Contrasting patterns of receptive field plasticity in the hippocampus and the entorhinal cortex: an adaptive filtering approach. Journal of Neuroscience 2002;22:3817-30.
  12. Frank LM, Brown EN, Wilson MA. Entorhinal place cells: trajectory encoding. In: Sharp P ed. Neural Basis for Navigation: Evidence from Single Cell Recordings, San Diego,         Kluwer, 2002; pp.97-116.
  13. Barbieri R, Quirk MC, Frank LM, Wilson MA, Brown EN. Construction and analysis of non-Poisson stimulus response models of neural spike train activity. Journal of Neuroscience Methods 2001;105:25-37.
  14. Frank LM, Brown EN, Wilson MA. A comparison of the firing properties of putative excitatory and inhibitory neurons from CA1 and the entorhinal cortex of the awake behaving rat. Journal of Neurophysiology 2001. Oct;86(4):2029-40.
  15. Brown EN, Nguyen DP, Frank LM, Wilson MA, Solo V. An analysis of neural receptive field plasticity by point process adaptive filtering. Proceedings of the National Academy of Sciences 2001; 98:12261-12266.
  16. Frank LM, Brown EN, Wilson MA. Trajectory encoding in the hippocampus and entorhinal cortex. Neuron 2000;27:169-178.
  17. Brown EN, Frank LM, Tang D, Quirk MC, Wilson MA. A statistical paradigm for neural spike train decoding applied to position prediction from ensemble firing patterns of rat hippocampal place cells, Journal of Neuroscience 1998;18:7411-7425.