|
|
Uri Tzvi Eden, Ph.D. Neuroscience Statistics Research
Laboratory
|
|
tzvi@neurostat.mgh.harvard.edu (617) 724-1060 |
Interests
My research
focuses on developing mathematical and statistical methods to analyze neural
spiking activity. I have worked to
integrate methodologies related to model identification, statistical inference,
signal processing, and stochastic estimation and control, and expand these
methodologies to incorporate point process observation models, making them more
appropriate for modeling the dynamics of neural systems observed through spike
train data. This research can be
divided into two categories; first, a methodological component, focused on
developing a statistical framework for relating neural activity to biological
and behavioral signals and developing estimation algorithms, goodness-of-fit
analyses, and mathematical theory that can be applied to any neural spiking
system; second, an application component, wherein these methods are applied to
spiking observations in real neural systems to dynamically model the spiking
properties of individual neurons, to characterize how ensembles maintain
representations of associated biological and behavioral signals, and to
reconstruct these signals in real time.
This research has been carried out at the Massachusetts General Hospital
within the Neuroscience Statistics Research Laboratory under the guidance of
Dr. Emery Brown.
Developing Mathematical Algorithms for
Neural Estimation.
The methodological component of my research
deals with the construction of a state-space framework for characterizing
random biological signals and relating them to the spiking activity of an
ensemble of neurons. As part of this
work, we have constructed discrete-time linear estimation algorithms that use
point process observations to estimate a state and provide confidence intervals
for those estimates (Eden et al., 2004).
This research makes explicit the relation between point process filters
and well-studied estimation algorithms for Gaussian observations such as the
Kalman Filter. It also provides
numerous extensions and additional estimation algorithms including an analogue
of a RLS estimation algorithm for static neural model parameters, a point
process smoother, mixed observation filters that combine information from
spiking observations with continuous valued signals such as those observed from
local field potentials (LFPs), and numerical methods to compute posterior state
distributions to arbitrary precision.
Additionally, we have been working on developing improvements to
particle filtering algorithms that take advantage of the localized nature of
neural spiking observations to improve the computational efficiency of these
numerical methods (Eden 2005; Ergun et al., 2005). Although the majority of this theoretical work deals with
discrete time estimation algorithms, we have also derived and analyzed the
theoretical properties of continuous time filters (Eden, 2005). The power of this methodological approach is
that it can be applied generally to any neural system whose firing properties
can be related to other biological and behavioral signals.
Characterizing Place Field Plasticity.
In collaboration with Dr. Loren Frank, we
have been investigating the problem of tracking and characterizing place field
plasticity in the rodent hippocampus.
We constructed dynamic neural models that related the spiking of
pyramidal cells in the CA1 region of hippocampus and in the deep entorhinal
cortex to the location of the animal and intrinsic temporal properties of the
neuron to capture bursting behavior and theta rhythmicity. Our research has shown that the past spiking
activity of neurons in these regions contributed substantially to our ability
to predict future spiking activity and that with repeated exposures to an
environment, neurons in both regions changed their firing properties in
systematic ways (Frank et al., 2002).
Decoding Reaching Movements from a
Dynamic Population of MI Neurons.
In collaboration with Dr. John Donoghue and
Dr. Wilson Truccolo at Brown University, we have been analyzing the firing
properties of neurons in primate primary motor cortex (MI) in relation to the
kinematics of arm movements and the neurons’ intrinsic and ensemble firing
history (Truccolo et al., 2004). We
have been interested in the problem of estimating intended reaching movements
using ensemble spiking observations, especially in the case where the
population of observed neurons and their firing properties are continually
changing (Eden et al., 2004). This
problem will be essential in the design of a chronically implantable motor
neural prosthetic device, as both the population of neurons that can be
observed and the firing properties of those neurons change on a daily
basis. We have also improved our state
space estimation paradigm to combine kinematic information from MI with
prescient information about reaching targets from motor planning regions
(Srinivasan et al., 2005)
Tracking Trial-by-trial Variability in
Learning Tasks.
In collaboration with Dr. Wendy Suzuki, Dr.
Silvia Wirth, and Dr. Eric Hargreaves at NYU, and with Dr. Gabriela Czanner at
the Neuroscience Statistics Research Laboratory, we have worked to explore the
dynamic behavior of neurons in the primate hippocampus while performing visual
associative memory tasks. Through this
research, we have developed new methods for characterizing the spiking
properties of neurons in repeated stimulus experiments. (Czanner et al.,
2005)
Characterizing and Tracking Learning
State
In collaboration with Dr. Wendy Suzuki, Dr.
Silvia Wirth, and Dr. Eric Hargreaves at NYU, and with Dr. Gabriela Czanner at
the Neuroscience Statistics Research Laboratory, we have developed new state
space estimation algorithms that use Bernoulli observations about success and
failure in combination with continuous valued observations about trial time to
estimate a learning state (Prerau et al., 2005) We have also developed methods that allow us to incorporate
information from spiking observations in our learning state model.
Characterizing Aberrant Oscillatory
Spiking Behavior in STN in Parkinson’s disease.
In collaboration with Dr. Emad Eskandar and
Dr. Ramin Amirnovin at MGH, we have been studying the spiking properties of
neurons in the subthalamic nucleus of humans with Parkinson’s disease before
and during a volitional movement task.
We have constructed simple generalized linear models that capture
stimulus related spiking properties as well as the effect of past spiking
history at short and long time scales.
We have found that these neurons have oscillatory firing patterns
whereby after a spike or burst is fired there is a wave of inhibition at 20-40
ms, followed by an increased probability of spiking at 40-90 ms. This firing pattern is attenuated when
planning and executing the arm movements.
We are currently preparing this work for publication.
Future
Research Directions
My research focus for the near future will
be to continue to develop the neural estimation paradigm theoretically and to
maintain my current collaborations and build upon these current studies, while
fostering new collaborations with experimentalists working in other neural
systems. Since the methodologies we
have developed are generally applicable to any neural system where the
observations include spike recordings, there are many opportunities to bring
together research from different experimental paradigms, multiple brain
regions, and even across different species under a common methodological
framework. The types of methodological
and system questions that I will actively be seeking experimental collaborators
to answer include: 1) relating the firing properties of neurons in multiple
connected brain regions, such as motor planning regions and MI, and using
combined observations from these regions to estimate biological and behavioral
signals; 2) combining multimodal neural data from both spike train recordings
and continuous valued recordings such as LFP or EEG data; 3) constructing
neural intensity models for new systems and relating these models to putative
deterministic models based on firing rates, where applicable; 4) studying
common statistical features of spiking activity that appear in multiple brain
systems.
Here is my complete curriculum vitae
Publications
Original Reports
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. J. Neurosci. 2002;22:3817-30
Eden UT, Frank LM, Solo V, Brown,
EN. Dynamic analyses of neural encoding
by point process adaptive filtering. Neural
Computation, 2004, 16(5), 971-998
Truccolo W, Eden
UT, Fellows MR, Donoghue JP, Brown EN.
Point process models for MI spiking activity: neural encoding and
decoding. J. Neurophysiology, 2004, 93:1074-1089.
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. Submitted, 2004
Srinivasan L, Eden UT, Willsky AS, Brown EN.
Minimal Probabilistic Description of Goal-Directed Movements. Submitted, 2005
Eden UT. Point process filters in the analysis of
neural spiking models. PhD. Thesis in
Medical Engineering/Medical Physics.
Harvard/MIT Division of Health Sciences and Technology. 2005
Prerau MJ, Eden, UT, Smith AC, Yanike
M, Suzuki WA & Brown EN. A mixed filter algorithm for simultaneously
recorded continuous-valued and binary observations. IEEE Transactions on
Biomedical Engineering, 2005, Submitted.
Proceedings of Meetings
Eden UT, Brown EN. Adaptive Filtering Algorithms for Spike
Train Observations. Poster presentation
at 10th Annual Computational
Neuroscience Meeting. San Francisco and
Pacific Grove, CA. June 30 – July 5,
2001
Eden UT, Smith A, Frank LN,
Barbieri R, Brown EN. Adaptive
filtering algorithms for neural encoding and decoding. Program No. 405.15. 2002 Abstract
Viewer/Itinerary Planner. Washington, DC: Society for Neuroscience, 2002.
Online.
Eden UT, Brown EN. Particle
Filtering Algorithms for Neural Decoding and Adaptive Estimation of Receptive
Field Plasticity. Poster presentation
at 11th Annual Computational Neuroscience Meeting. Chicago, IL. July 21-25, 2002
Eden UT, Truccolo W, Barbieri
R, Donoghue JP, Brown EN. Adaptive
Neural Filtering Applied to Hand Movement Coding in Primate Primary Motor
Cortex During a Hand Tracking Task.
Poster presentation at 12th Annual Computational Neuroscience
Meeting. Alicante, Spain. July 5-9,
2003
Eden UT,
Truccolo W, Ergun, A, Fellows MR, Donoghue JP, Brown EN. Exact and approximate point process filters
for adaptive neural encoding and decoding.
Program No. 429.2. 2003 Abstract
Viewer/Itinerary Planner. Washington, DC: Society for Neuroscience, 2003.
Online.
Eden UT, Brown EN. Adaptive
Decoding of Hand Movement Trajectories from Simulated Spike Train Observations
from a Dynamic Ensemble of Motor Cortical Neurons. Poster & Oral presentation at 13th Annual
Computational Neuroscience Meeting & Workshops. Baltimore, MD. July 18-20, 2004
Truccolo W, Fellows MR, Eden
UT, Brown EN, Donoghue JP. Primary
motor (MI) and parietal (5d) coordination during reaching: point process and
LFP models. Program No. 421.1. 2004 Abstract Viewer/Itinerary Planner.
Washington, DC: Society for Neuroscience, 2004. Online.
Eden UT, Truccolo W, Fellows MR,
Donoghue JP, Brown EN. Reconstruction of Hand
Movement Trajectories from a Dynamic Ensemble of Spiking Motor Cortical
Neurons. Oral presentation at 26th
annual international conference IEEE engineering in medicine and biology
society. San Francisco, CA. Sept. 1-5, 2004
Eden UT, Brown EN. Using dynamic algorithms to decipher neural
representations of biological signals.
Oral presentation at AMS Special
Session on Mathematics and 21st Century Biology, Joint Mathematics
Meetings. Atlanta, Georgia. January 5,
2005
Srinivasan L, Eden UT, Willsky AS, Brown
EN. Goal-directed state equation for
tracking reaching movements using neural signals. Proceedings of the 2nd International IEEE EMBS
Conference on Neural Engineering, Arlington VA, March 2005
Czanner G, Eden UT, Wirth S, Suzuki WA, Brown
EN. A dynamic analysis of neuronal
spiking activity in the primate hippocampus.
Program No. 776.4. 2005 Abstract Viewer/Itinerary Planner.
Washington, DC: Society for Neuroscience, 2005. Online.
Textbook Chapters
Brown EN, Barbieri R, Eden UT, Frank LM. Likelihood methods for neural spike train
data analysis, In: Computational neuroscience: a comprehensive approach.
London, CRC Press. 2003; Chapter 9, pp 253-286
Thesis
Eden
UT.
Point process filters in the analysis of neural spiking models. PhD. Thesis in Medical Engineering/Medical
Physics. Harvard/MIT Division of Health
Sciences and Technology. 2005
Patents
Srinivasan
L, Eden UT, Brown EN & Willsky A. Device and method for providing a
combined bioprosthetic specification of goal state and path of states to goal.
Filed, January 27, 2005.
Eden UT & Hickerson K.
Accelerated handwritten symbol recognition in a pen based tablet
computer. Filed, May 3, 2001.
Eden UT & Eden G. Method for preventing dehydration from
alcohol ingestion. Filed August, 2005