oscillations in mammalian sensorimotor processing
DESCRIPTION
Oscillations in Mammalian Sensorimotor Processing. Diane Whitmer Dissertation Defense Division of Biological Sciences September 29, 2008. Overview. Oscillations in the Rat Vibrissa System A. How can rats use their whiskers to locate objects? B. Does hippocampal theta drive whisking? - PowerPoint PPT PresentationTRANSCRIPT
Oscillations in Mammalian Sensorimotor
Processing
Diane WhitmerDissertation Defense
Division of Biological SciencesSeptember 29, 2008
Overview
I. Oscillations in the Rat Vibrissa SystemA. How can rats use their whiskers to locate objects?B. Does hippocampal theta drive whisking?
II. Visually cued Finger Movements in Human Epilepsy PatientsA. What is the neural signature of movements?B. (How) Should intracranial signals be un-mixed? **ICA**
III. Conclusions and Next Steps
Overview
I. Oscillations in the Rat Vibrissa SystemA. How can rats use their whiskers to locate objects?
B. Does hippocampal theta drive whisking?
II. Visually cued Finger Movements in Human Epilepsy PatientsA. What is the neural signature of movements?B. (How) Should intracranial signals be un-mixed?
III. Conclusions and Next Steps
Rats engage in exploratory
“whisking”
Berg & Kleinfeld, 2003
The Rat Vibrissae Pathway
Kolb and Tees, 1990
Deschenes et al., 2001
Brecht et al., 1997
Mehta, Whitmer et al., 2007
Coding strategies for object localization
Small or no movements
Mehta, Whitmer et al., 2007
Coding strategies for object localization
Small or no movements
Whisker movements
Behavioral testing of whisking
Lever
Water Fountain
Position Sensor
Reward Outlet Vacuum
Reward Inlet Valve
Nose Sensor
Restraint Bar
Mehta, Whitmer et al., 2007
Mehta, Whitmer et al., 2007
Responses from a Testing Session
Mehta, Whitmer et al., 2007
Mehta, Whitmer et al., 2007
Responses Latencies
Mehta, Whitmer et al., 2007
Coding strategies for object localization
Significance for the Rat Vibrissa System
• Vibrissa process sensory information about What and Where
• Results from discrimination of location in rostral-caudal plane suggests overall scheme for position in 3-d space:
Ahissar & Knutsen, 2008
Overview
I. Oscillations in the Rat Vibrissa SystemA. How can rats use their whiskers to locate objects? Information about the location of the whisker is combined with contact information.B. Does hippocampal theta drive whisking?
II. Visually cued Finger Movements in Human Epilepsy PatientsA. What is the neural signature of movements?B. (How) Should intracranial signals be un-mixed?
III. Conclusions and Next Steps
Overview
I. Oscillations in the Rat Vibrissa SystemA. What is the significance of phase in the whisking cycle?Information about the location of the whisker is combined with contact information.B. Does hippocampal theta drive whisking?
II. Visually cued Finger Movements in Human Epilepsy PatientsA. What is the neural signature of movements?B. (How) Should intracranial signals be un-mixed?
III. Conclusions and Next Steps
Hippocampal theta rhythm is associated with voluntary movement in the rat
• running
• jumping
• exploratory head movements
• swimming
Vanderwolf, 1969
Berg, Whitmer, Kleinfeld, 2006
Are these two signals phase-locked?
Coherence quantifies phase-locking
Berg, Whitmer, Kleinfeld, 2006
Coherence quantifies phase-locking
Berg, Whitmer, Kleinfeld, 2006
Trial to trial variability of coherence between whisking and hippocampal theta
Berg, Whitmer, Kleinfeld, 2006
Berg, Whitmer, Kleinfeld, 2006
Trial to trial variability of coherence between whisking and hippocampal theta
Berg, Whitmer, Kleinfeld, 2006
Coherence between whisking and hippocampal theta is not significant
Overview
I. Oscillations in the Rat Vibrissa System A. What is the significance of phase in the whisking cycle?Information about the location of the whisker is combined with contact information. B. Does hippocampal theta drive whisking? NO
II. Visually cued Finger Movements in Human Epilepsy PatientsA. What is the neural signature of movements?B. (How) Should intracranial signals be un-mixed?
III. Conclusions and Next Steps
Overview
I. Oscillations in the Rat Vibrissa SystemA. What is the significance of phase in the whisking cycle?Information about the location of the whisker is combined with contact information.B. Does hippocampal theta drive whisking? NO.
II. Visually cued Finger Movements in Human Epilepsy PatientsA. What is the neural signature of movements?B. (How) Should intracranial signals be un-mixed?
III. Conclusions and Next Steps
Scales of measurement of
electrophysiological brain signals
Churchland & Sejnowski, 1992
Electroencephalography (EEG) recordings
Jasper & Penfield, 1949
Cocktail Party
Independent Component Analysis of EEG Data
Makeig, Bell, Jung & Sejnowski, 1996
CSF
EEG
Assume that sources are:
1. Statistically independent
2. Volume conduction instantaneous (no time delays)
3. Sources mix linearly to produce channel data
4. Spatially stationary
Independent Component Analysis
x = A s (theory) x: recorded channel data
s: actual underlying sources
A: “mixing matrix”
Onton & Makeig, 2006
1. ICA separates EEG data into different brain rhythms that are modulated during this working memory task.
Voltage
Onton & Makeig, 2006
2. The maps from projecting independent components onto the electrodes produce biologically plausible patterns (dipoles)
CSFEEG
??
Standard EEG ICA of EEG Intracranial EEG (iEEG)
ICA of iEEG
Is ICA useful for the interpretation of intracranial data?
Three Ways to Assess:
1. Are the time series of intracranial channels statistically independent? (Control)
2. Do independent component maps appear consistent with anatomically and/or functionally linked brain regions?
3. Does ICA separate functionally distinct brain processes?a. Pathological signals?
b. Event-related dynamics?
Patient Electrode Locations
Intracranial Montage:
Right Lateral hemisphere
6x8 Grid
Two 8-contact strip
Right mesial surface
Three 4-contact strips
Two 4-contact strips
One 4-contact strip
Inter-hemispheric fissure
Mesial temporal lobe
Orbital Frontal Surface
Right frontal & lateral lobe
Lateral temporal lobe
right ring
Time (msec)
Stimulus Key-press Beep
0
Next Stimulus
ISI: 1.570 sec
Visually Cued Finger Movement Task
Task Design
10 Trials Per Finger Per Condition (N = 400)
Block Design:L pic - R pic - L word - R wordL pic - R pic - L word - R word
Finger presentation randomized within a block
epoch: 2 sec
Grid 24
Grid 25
Grid 26
Grid 27
Grid 28
Grid 29
IC1
IC2
IC3
IC4
IC5
IC6
Example Channel (black) and Component (blue) Time Series
ICA of Intracranial Data
ICA of Intracranial Data
Reduction in pairwise mutual information from channels to components
ICA of Intracranial Data
Reduction in pairwise mutual information from channels to components
Independent component map are consistent with anatomically and/or functionally
linked brain regions
Focal
Diffuse
Complex
Right lateral frontal Grid
Lateral Temporal Strips Mesial Temporal Strips
Orbital Frontal Surface strip
Strips in anterior frontal interhemispheric fissure
= FIRDA: frontal intermittent rhythmic delta, reportedly synchronous
ICA separates pathological “FIRDA” acivity
Epileptic “Frontal Intermittent Rhythmic Delta Activity” (FIRDA)
IC3
Cortical signatures of movement
Jasper & Penfield, 1949
Jasper & Andrews, 1936
Cortical signatures of movement
Jasper & Penfield, 1949
Miller et al, 2007
Jasper & Andrews, 1936
Alpha/beta power
decrease
Cortical signatures of movement
Jasper & Penfield, 1949
Miller et al, 2007
Gamma power increase
Alpha/beta power
decrease
Jasper & Andrews, 1936
Finger movementFinger movement
Finger movement
Mu blocking on Grid24
Grid24 log spectral power
Finger movement
IC18 captures classic event-related spectral changes and mu blocking associated with finger movement
IC18, 89% of Grid24
Finger movement
Finger movement
ICA finds components with classic movement-related dynamics
Independent components identify components in overlapping brain areas with different dynamics
ICA is useful for the interpretation of intracranial data
Three Ways to Assess:
1. ICA finds a set of time series that are more statistically independent than the sensor data (sanity check)
2. Independent component maps appear consistent with anatomically linked brain regions
3. ICA separates functionally distinct brain processes:a. Pathological signals
b. Event-related dynamics
Next Steps for ICA of Intracranial Data
1. ICA for the interpretation of cognitive task data for which the dynamics are not known in advance
Next Steps for ICA of Intracranial Data
1. ICA for the interpretation of cognitive task data for which the dynamics are not known in advance
2. Advanced ICA methods- Complex/convolutive ICA- Multiple mixtures ICA
Next Steps for ICA of Intracranial Data
1. ICA for the interpretation of cognitive task data for which the dynamics are not known in advance
2. Advanced ICA methods- Complex/convolutive ICA- Multiple mixtures ICA
3. Source localization: patient-specific “forward model” that accounts for craniotomy
Oscillations
1. Whisker movements are oscillatory, and can be used to locate objects in space.
2. Hippocampal theta is not the rhythm that drives vibrissa movements.
3. Coherence can be used to determine whether oscillations are phase-locked.
4. Alpha, beta, and gamma oscillations correspond to voluntary movements in the human.
Soloman & Hartmann, 2006
Robotic Whiskers
Prosthetic Arm
Tetraplegic patient controls computer cursor with brain signals
photo from Donoghue Lab
Applications to Brain-Computer Interfaces
Caplan et al., 2006
Acknowledgements
Thesis Committee
Terry Sejnowski
David Kleinfeld
Scott Makeig
Greg Worrell
Eric Halgren
Pam Reinagel
Gert Cauwenberghs
Woods Hole
Neuroinformatics
Partha Mitra
Hemant Bokil
Ifije Ohiorhenuan
Jean Verrette
Swartz Center
Zeynep Akalin-Acar
Bob Buffington
Arno Delorme
JR Duann
Toby Fernsler
Klaus Gramann
T-P Jung
Il Keun Lee
Ryan Low
Julie Onton
Jason Palmer
Nima Bigdely Shamlo
Elke Van Erp
Andrey Vankov
Ying Wu
Kleinfeld Lab
Rune Berg
Omar Clay
Dan Hill
Rodolfo Figueroa
Samar Mehta
Quoc Nguyen
Nozomi Nishimura
Chris Schaffer
Lee Schroeder
Phil Tsai
Mayo Clinic
Matt Stead
Karla Crockett
Acknowledgements
Thesis Committee
Terry Sejnowski
David Kleinfeld
Scott Makeig
Greg Worrell
Eric Halgren
Pam Reinagel
Gert Cauwenberghs
Woods Hole
Neuroinformatics
Partha Mitra
Hemant Bokil
Ifije Ohiorhenuan
Jean Verrette
Swartz Center
Zeynep Akalin-Acar
Bob Buffington
Arno Delorme
JR Duann
Toby Fernsler
Klaus Gramann
T-P Jung
Il Keun Lee
Ryan Low
Julie Onton
Jason Palmer
Nima Bigdely Shamlo
Elke Van Erp
Andrey Vankov
Ying Wu
Kleinfeld Lab
Rune Berg
Omar Clay
Dan Hill
Rodolfo Figueroa
Samar Mehta
Quoc Nguyen
Nozomi Nishimura
Chris Schaffer
Lee Schroeder
Phil Tsai
Mayo Clinic
Matt Stead
Karla Crockett
Acknowledgements
Thesis Committee
Terry Sejnowski
David Kleinfeld
Scott Makeig
Greg Worrell
Eric Halgren
Pam Reinagel
Gert Cauwenberghs
Woods Hole
Neuroinformatics
Partha Mitra
Hemant Bokil
Ifije Ohiorhenuan
Jean Verrette
Swartz Center
Zeynep Akalin-Acar
Bob Buffington
Arno Delorme
JR Duann
Toby Fernsler
Klaus Gramann
T-P Jung
Il Keun Lee
Ryan Low
Julie Onton
Jason Palmer
Nima Bigdely Shamlo
Elke Van Erp
Andrey Vankov
Ying Wu
Kleinfeld Lab
Rune Berg
Omar Clay
Dan Hill
Rodolfo Figueroa
Samar Mehta
Quoc Nguyen
Nozomi Nishimura
Chris Schaffer
Lee Schroeder
Phil Tsai
Mayo Clinic
Matt Stead
Karla Crockett
Acknowledgements
Thesis Committee
Terry Sejnowski
David Kleinfeld
Scott Makeig
Greg Worrell
Eric Halgren
Pam Reinagel
Gert Cauwenberghs
Woods Hole
Neuroinformatics
Partha Mitra
Hemant Bokil
Ifije Ohiorhenuan
Jean Verrette
Swartz Center
Zeynep Akalin-Acar
Bob Buffington
Arno Delorme
JR Duann
Toby Fernsler
Klaus Gramann
T-P Jung
Il Keun Lee
Ryan Low
Julie Onton
Jason Palmer
Nima Bigdely Shamlo
Elke Van Erp
Andrey Vankov
Ying Wu
Kleinfeld Lab
Rune Berg
Omar Clay
Dan Hill
Rodolfo Figueroa
Samar Mehta
Quoc Nguyen
Nozomi Nishimura
Chris Schaffer
Lee Schroeder
Phil Tsai
Mayo Clinic
Matt Stead
Karla Crockett
Acknowledgements
Family
Priscilla, Ellen, Roger, Rachel, Ralf, Julian, Jonas, Jim, Lisa
Friends
Zoe Argento, Alicia Bicknell, Rael Cahn, Kim Ditomasso, Gloriana
Gallegos, Andra Ghent, Jody Harrell, Kaori Hirata, Dan Keller, Mina
Kinukawa, Jessica Kleiss, Debra Knight, Zoe Langsten, Oanh Nguyen,
Fij Ohiorhenuan, Steve Oldenburg, Jeff Slattery, Ben Sullivan, Emilija
Simic, Corinne Teeter, Elke Van Erp, Jean Verrette, Shane Walker,
Amaya Becvar Weddle
Acknowledgements
Family
Priscilla, Ellen, Roger, Rachel, Ralf, Julian, Jonas, Jim, Lisa
Friends
Zoe Argento, Alicia Bicknell, Rael Cahn, Kim Ditomasso, Gloriana
Gallegos, Andra Ghent, Jody Harrell, Kaori Hirata, Dan Keller, Mina
Kinukawa, Jessica Kleiss, Debra Knight, Zoe Langsten, Oanh Nguyen,
Fij Ohiorhenuan, Steve Oldenburg, Jeff Slattery, Ben Sullivan, Emilija
Simic, Corinne Teeter, Elke Van Erp, Jean Verrette, Shane Walker,
Amaya Becvar Weddle
My dissertation is dedicated to the lovingmemory of my grandfather Martin Littman,who wanted to celebrate this day but passedaway on March 16, 2006.
Additional Slides
Traveling waves in cortex
Ermentrout & Kleinfeld, 2001
Traveling waves in cortex could appear synchronous from a distance when averaged
over sizable cortical patch
1.5 cm
Example
10 Hz wave
velocity ~= 200 cm/sec
2pi * 10 Hz * 1.5 cm / 200 cm/sec
= 0.15pi = 27 degrees
27 degree phase difference between center and edge
3 cm cortical patch
Traveling waves in primate motor cortex
Example
40 Hz wave
velocity = 28 cm/sec
2pi * 40 Hz * 0.2 cm / 28 cm/sec
= 0.57pi = ~102 degrees
Sizable phase difference between center and edge of the electrode array.
Do the waves travel the entire distance of an area of cortex over which the activity would be averaged by iEEG?
0.4 cm width of electrode array
Rubino et al., 2006
Principle Component Analysis (PCA) versus Independent Component Analysis (ICA)
PCA ICA
Based on variance Based on statistical independence (stronger requirement)
Groups together the “sources” of the data
Separates the data sources
Typically used for dimensionality reduction
Requires separate method for dimensionality reduction (e.g. PCA, then ICA)
Infomax ICA Algorithm
Define: x(t) = A*s(t)
Goal: find u and W such that W*x(t) = u(t)
and u is independent: p(u) = p1(u1)*p2(u2)*...pN(uN)
1. Sphere the data: diagonalize the covariance matrix of x: <xxT> = I
2. Maximize the joint entropy of Y = g(u), where g is sigmoid
3. Find a matrix W such that max{H(g(Wx))}
4. Define a surface H(g(Wx))
5. Find the gradient d/dW H(...) and ascend it
6. When the gradient is zero, a maximum is reached
Applications of ICA
• Acoustics: cancellation of acoustic reverberations
• Geophysics: seismic deconvolution
• Image processing: restoration of images
Theta phase for encoding spatial location
Buzsaki, 2004
Latency Distribution
Mehta, Whitmer et al., 2007
Mehta, Whitmer et al., 2007
ROC curve
Mehta, Whitmer et al., 2007
Controls
Berg, Whitmer, Kleinfeld, 2006
Peak amplitudes are not correlated
|C |= P1/(NK −1)
Confidence limits on coherence estimates
Reduction in pairwise mutual information from channels to components
ICA of Intracranial Data
Role of beta oscillations in motor system
• Preparatory motor activity (Sanes & Donoghue)
• Maintain steady contractions of contralateral muscles
• Bind sensory and motor areas during motor maintenance behavior (Brovelli et al., 2003)
• Priming of motor movement for receiving sensory input
• Clock for coordinating timing of movements
The Scientific Method