(navigation image)
Home American Libraries | Canadian Libraries | Universal Library | Community Texts | Project Gutenberg | Children's Library | Biodiversity Heritage Library | Additional Collections
Search: Advanced Search
Anonymous User (login or join us)
Upload
See other formats

Full text of "Fundacion Mas i Manjon - Foundation's Mas & Manjon -Since 1975-"

REVISION MAIG 2008 FUNDACION J.MAS 

Neural Modeling and Functional 
Brain Imaging: An Overview 



Barry Horwitz 
Brain Imaging & Modeling Section 

NIDCD, NIH 



Methods to Understand Neural Basis 

of Human Cognition 



1. Brain lesions & cognitive neuropsychology 

2. Electrophysiological recordings in primates (mammals) 

3. Pharmacological and genetic studies 

4. Transcranial magnetic stimulation 

5. Functional neuroimaging 

Hemodynamic-metabolic methods (PET, fMRI) 
Electric-magnetic methods (ERP, MEG) 



All these data are generally 
incommensurate with one another. 



Methods to Understand Neural Basis 

of Human Cognition 



1. Brain lesions & cognitive neuropsychology 

2. Electrophysiological recordings in primates (mammals) 

3. Pharmacological and genetic studies 

4. Transcranial magnetic stimulation 

5. Functional neuroimaging 

Hemodynamic-metabolic methods (PET, fMRI) 
Electric-magnetic methods (ERP, MEG) 

Functional neuroimaging enables the activity of all brain 
regions to be seen simultaneously - hence, network 

analysis becomes necessary. 



Functional Neuroimaging Methods 



Two-dimensional surface imaging 

(e.g., 133-Xenon inhalation method) 
Optical imaging 
Single photon emission computed tomography 

(SPECT) 
Positron emission tomography (PET) 
Functional magnetic resonance imaging (fMRI) 
Evoked potentials, electroencephalography (ERP,EEG) 
Magnetoencephalography (MEG) 



Functional Neuroimaging Methods 



Two-dimensional surface imaging 

(e.g., 133-Xenon inhalation method) 
Optical imaging 
Single photon emission computed tomography 

(SPECT) 
Positron emission tomography (PET) 
Functional magnetic resonance imaging (fMRI) 
Evoked potentials, electroencephalography (ERP,EEG) 
Magnetoencephalography (MEG) 
Transcranial magnetic stimulation (TMS) 



MCd^irr^Tr^rffTdnr phantafie 



P^eani 




I /rlr Century anonymous engraving. Tho Physician Curing Fantasy. By permission. Philadelphia Museum of Art. 



Relation of Brain Functional 
Activity to Cerebral Blood Flow 



Brain functional activity 

Cellular work 

Cerebral oxidative metabolism 

Delivery of oxygen and glucose 
Cerebral blood flow (CBF) 







Characteristics: 
Functional Neuroimaging Methods 




Tec! 


inique 


i Variable 


Time Resol. 


Spatial Resol. 


Comments 


2D -Surface 
Imaging 


1 33-Xe 
rCBF 


5-7 min 


2.5 cm 


nontomographic 
no deep structures 
3 runs 


SPECT 


99m-Tc-HMPAO/rCBF 1-2 min 
99m-Tc-ECD/rCBF 1-2 min 
133-Xe/rCBF 2 min 


7 mm 
7 mm 
7 mm 


semiquant.;2 runs 

2 runs 

3 runs 


PET 


[15-0]water/rCBF 
[18-F]FDG/rCMRglc 


15sec-1 min 
15-30 min 


5 mm 
5 mm 


multiple runs 
1 run 


ERP 


scalp-recorded 
electric potetial 


1 msec 


Very coarse 


many trials 


MEG 


magnetic field/ 
current source 


1 msec 


A few mm 


many trials 
nonuniquesoln. 


fMRI 


blood oxygenation 


A few sec. 


2 mm 


many runs 
semiquant. 
source of signal 



Positron Emission 
Tomography 



Subtraction of rCBF Images 



mill QQgi mitt 
175 



f* 






tVT 



i^ 



Face Matching Sensorimotor Matching - Control 
Control 



[IHILATIGN 
VINT 



DETECTOR 
RINGS 



*■ 



M!TTED £*= 

hc/tons -^ 



COMPUTER AND 
INTERFACES + ♦ 



♦♦♦♦♦♦♦♦♦♦♦♦ 



f|«' 



! 



frequency 
coil 

Prism glasses 



fMRI Setup 




imulus control 
computer 



WMWMnKK&mnmwi, 



Hi'lnlH~i 



control computer 



Jody Culham - fMRI Course (http://defiant.ssc.uwo.ca/Jody_web/courses.htm) 



Blood Oxygenation Level Dependent (BOLD) Signal 



REST 



Hb0 2 



Magnetic Field Lines 



ACTIVATION 



^O * - Hb Q„ 



Vfc>0*_T 



Magnetic Field Lines 



Blocked vs. Event-related 



BLOCKED: 



SPACED MIXED TRIAL: 



RAPID MIXED TRIAL: 

luiniuiniuiiiiinnnnfuinii fum FJininuuiniuuinnnn 

k kkkkkkkkkkkkkkkkkkkk- kkk k k kk 



Jody Culham - fMRI Course 



Source: Buckner 1998 



Block Designs 



BLOCKED: 



= trial of one type 
(e.g., face image) 



= trial of another type 
(e.g., place image) 




j-,i -■■ .;.. y 



Assumption: Because the hemodynamic response delays and blurs 
the response to activation, the temporal resolution of fMRI is limited. 



" 2 + 

3 i 

O 

to 



Positive 
BOLD response 




Post stimulus 
undershoot 

-A,,,, , 



Stimulus 



Jody Culham 
- fMRI Course 



Statistical Maps & Time Courses 



I A a4.colorobjscraml.fmr 



Use stat maps to pick regions 
Then extract the time course 



|4S* ROI Signal Time C( 



Jody Culham - fMRI Course 



What are the temporal limits? 

What is the briefest stimulus that fMRI can detect? 
Blamire et al. (1992) - 2 sec 
Bandettini (1993): 0.5 sec 
Savoy et al (1995): 34 msec 



HW1R! BOLD SIGNAL TO PULSED VISUAL STIMULATION 



2.0 - 



LU 

i 

U 



LU 



^1000 
msec 




TIME (SCOWS) 



Jody Culham - fMRI Course 



MEG Study of Visual Word Processing 




-v- 






-Aw"^, 



-v*— -JWA- jV^u* -^u. 
yU^ ^A«„ WW- -V 1 - -v*** • JW ^' *«' W ' rT*" 

— h ^~ 'V^ -i/^Wc ^V^-> 



•V 41 " -^V""* 






~J»<* 



w^/^jiii! 



•*\Ni^ V*v*- 












50fT/cm L 






4^ 




400 800 ms 



200 ms 



-Y*v-* 



-yr' 







Figure 3 



Acquiring and Interpreting the Signal 



Relation of neural activity to the variable that is to 
be imaged (e.g., rCBF, blood oxygenation). 

Relation of the imaged variable to the the signal 
(e.g., brain radioactivity, intensity in T2*-sensitized 
images). 

Physics of measuring device (e.g., photodetectors, 
RF coils). 

Hemodynamics/metabolism of labeled quantity 
(e.g., [18-F]FDG, oxygenated hemoglobin). 

Paradigm design. 

Statistical analysis of image. 

Interpreting the results in terms of 
neurobiology/cognitive neuroscience. 



Neurobiological Substrate of 
Functional Neuroimaging Signals 



What is the relation between neuronal activity (ion flux) and 
vascular/metabolic response? Some information will come from 
from optical imaging studies, some from fMRI analyses, some 
from nonhuman animal rCBF experiments. 

Need to scale up this relation to the level of a cortical column or 
other well-defined anatomical structure. Need to determine the 
role neuromodulatory transmitters (e.g., ACh, 5HT, etc.) play in 
regulating rCBF. 

This method was used to understand some information about 
the neurobiological source of ERP. 

One then can use modeling to assess the importance of the 
firing of various synaptic populations on the functional 
neuroimaging signal. 



Some questions That Computational 
Neuroscientistists Can Address With 
Respect To Functional Neuroimaging 



Biological Substrates of Neuroimaging Signals 

Relation of Dynamic Measures of Functional 
Activity (e.g., MEG) to "Steady-State" Measures 
(e.g., PET)* 

Systems-Level (Network) Modeling 

Relation of Systems-Level Models to Neuronal, 
Ensemble and Cognitive Models 



Research Areas 



Functional Brain 

Imaging 

Experiments 



Functional & 

Effective 

Connectivity 

Network Analysis 



Determine brain regions constituting 
hypothesized network mediating task 



Large-Scale 
Neurally 
Realistic 
Modeling 



Construct 
dynamic neural 
network model 
mediating task 



PET/fMRI Data Analysis Strategies 



Subtraction paradigm (Functional Segregation) 

Task changes neural activity in a region. 

Compare task with control task to find brain regions 
used by task. 

Covariance paradigm (Functional Integration) 

Task is mediated by a set of interacting brain regions, 

Regions whose activities are correlated are part of 
neural network mediating task. 



Broca's Area and Language Production 



Broca's area 

■ Left inferior frontal gyrus (LIFG) 

- Lesions result in a language production deficit 

■ Not well defined - many tasks activate this part of LIFG 

- Braun et al. (Brain, 2001) used PET to study speech and ASL 

Cytoarchitectonics 

■ Layering of neurons in the 6 
cortical layers 

■ Brodmann areas 

■ Probabilistic brain atlas 
Questions 

■ Do speech and ASL activate 
Broca's area (BA44 and 
BA45)? 

- Is Broca's area activated by 
nonlanguage tasks? 




Probabilistic Brain Atlas 



PET Activation 
[|Z| > 2.33] 



Periphery 
Region 
Activated 
[|Z| > 2.33] & 
[BA45<50%] 



Core Region 
Activated 
[|Z| > 2.33] & 
[BA45>50%] 




Core Region 
[BA45>50%] 



Periphery 

Region 

[BA45<50%] 



Brodmann Area 45 and Language Production 





~ i\ 


1 


is 






SPEECH vs. 
MOTOR Control 



ASLvs. LIMB 
Control 




SPEECH vs. 

MOTOR Control 

Monolinguals 
Conclusions: 

1. Similar neural substrates in BA45 are used by ASL and Speech. 

2. Non-language tasks can activate BA44 (not shown). 



Functional Connectivity 



Interregional functional 
connectivity is evaluated as 
the correlation coefficient 

(1) between subject-to-subject 
activities for PET 

or 

(2) between interregional time- 
series activities for fMRI. 



rCBF (left ang. gyrus) 



(Horwitz et al., PNAS, 1998) 



Results -- Subtraction Paradigm 



Tagamets et al. (J. Cogn. Neurosci., 
2000), using SPM, found a shift in the 
pattern of activity in going from words 
to pseudowords to letter strings to 
false fonts. 
Lateralization changes 

left to bilateral in posterior cortex 
left to right in frontal cortex 
Increased recruitment of parietal 

cortex 
Decreased activity in left post. temp. 

cortex 



Words 



Pseudowords 



llxttcr Suing; 



The above changes reflect shifts in the 
semantic and phonological content of 
the stimuli. 



False Fonts 



Functional Connectivity Map - Reference Voxel in BA44/45 (-50 28 16) 



# 



*fy 



S^/j 



^wr r wr r Y 



ft*Vw 



\ 



V 



w 






X 



^( 



N 



W 



K 



*< 



M 



M 



A 1 /\ 



/ ^ 



CZciltir Legend: words, pseudowords & letter-strings 
words & pseudowords 
words & letter-strings 






words 
pseudowords 

letter-strings 






(Bokde et al., Neuron, 2001) 



Functional Connectivity Map - Reference Voxel in BA47/10 (-48 36 -14) 



r dF r l > 



'iw/f \i>wy^v>* 



^VH 






T 



VV 



is^/d 






"P 



K 



K 



i 



¥ 



K 



)( 



i< 



*< 



P * 



p % 



M 



M 



A 



M 






,! 



I 



r 



J^S^^^^iaBJ^^^iJ^ 



Color Legend: words, pseudowords & letter-strings 
words & pseudowords 
words & letter-strings 



words 

pseudo words 
letter-strins?s 



Brain Connectivity 



Anatomical Connectivity 

Functional Connectivity 

Effective Connectivity 



Systems-Level Network Analysis 

(Structural Equation Modeling or Path Analysis) 



Correlations between regions may be due to both direct 
and indirect effects. 



Goal: account for the observed correlations between all regions 
of interest in terms of the known functional linkages (path 
coefficients) between these regions. 

Starting point: covariance or correlation matrix between activity 
in each region of interest, along with the anatomical connections 
between these regions. 



Method: functional strengths of these linkages are adjusted until 
calculated and observed correlation matrices are as identical as 
possible. 

Result: functional systems-level neural model corresponding to 
the task of interest. 



Path Analysis Models for Face an 
lot-Location Match in 



Face Matching 




Dot-Location Matching 




Path Coefficients 



Positive 

0.7 to 1.0 — 

0.4 to 0.6 

0.1 to 0.3 



Negative 

-0.7 to -1.0 
-0.4 to -0.6 
-0.1 to -0.3 






Mcintosh et al., J. Neuroscience, 1994 



Problems with Relating Hemodynamic 
Data to Underlying Neural Activity 



1. Spatial resolution - each PET or fMRI resolvable element 
contains multiple and diverse neuronal populations. 

2. Temporal resolution - temporal resolution of neuronal 
activity is on the order of milliseconds; PET and fMRI 
(because of hemodynamic delay) is on the order of 
seconds; fast transients may be invisible to PET/fMRI. 

3. Synaptic vs. neuronal activity - electrical activity comes 
from cell body firings, PET/fMRI reflect primarily the 
activity of synapses; excitatory vs. inhibitory. 

4. Connectivity - PET/fMRI activity is a mixture of local and 
afferent synaptic activity. 



Large-Scale Neural Modeling 



Goal: Construct a large-scale, neurobiologically 
realistic neural model that can perform tasks like 
those studied by PET and fMRI. 



Multiple, interconnected brain regions (feedforward and feedback 
connections). 

Each region consists of multiple neuronal units (cortical column). 

The basic unit consists of an excitatory-inhibitory pair. 

Model can perform multiple tasks (e.g., DMS for shape, control task). 

Dynamic behavior of excitatory units in each region matches that 
observed by primate electrophysiological studies. 

Synaptic activity (both excitatory and inhibitory), integrated spatially 
and temporally, represents rCBF/BOLD. 



Large-Scale Neural Modeling 



Visual Delayed-Match to Sample for Shape (Visual 
Object Processing): PET and fMRI (Tagamets & Horwitz, 

Cerebral Cortex, 1998; Horwitz & Tagamets, HBM, 1999) 

Transcranial Magnetic Stimulation (TMS) and PET 

(Husain et al., Neurolmage, 2002) 

Inhibition and PET/fMRI Activity (Tagamets & Horwitz, Brain 
Res. Bull., 2001) 

Auditory Delayed Match-to-Sample for Tonal 
Patterns (Auditory Object Processing) 

Perceptual Grouping for Auditory Objects 

Functional Connectivity: PET and fMRI 



Neuroanatomy for Visual and 
Auditory Object Processing 




4b Arcuate 



4b Medial 



4b Principalis 



PS 
open 
4bVentrj 



2 Beit 




3 Parabelt 



4a Temporal Kaas et aL 

Current Opinion in Neurobiology 



Much less is known about the neuroantomy and 
neurophysiology of primate auditory pathways than 
about the visual pathways. 



Delayed Matched-to-Sample Tasks 



Shape 



Tonal pattern 



Stimulus 1 Delay Stimulus 2 Response ITI, next trial 



TIME 



Regions of the Visual Model 



LGN (stimulus) 



V1/V2 



Prefrontal 



(Tagamets & Horwitz, Cerebral Cortex, 1998) 



Neuronal activity in Monkey PFC During an 

Oculomotor Delay Task 



A. C group 




B. C+D group 




C. D group 




20 3/3 



1 a 

Fig. 3. Composite histograms summing over a large r.umber of 
neurons recorded from the principal sulcus during the ODR 
task. Only trials for a neuron's preferred direction {largest 
response) are included. (A) Composite histogram of 27 neurons 
that responded to the cue. (B) Composite of 33 neurons that 
had both phasic cue-period activity and tonic delay-period ac- 
tivity. (C) Composite histogram of 78 neurons that exhibit only 
tonic delay-period activity. C = cue; D = delay; R = response 
periods. (From Funahashi et al., 1990a.) 



Basic Unit 
(Cortical column) 



Basic Unit of Model and 
Between-Area Connections 



60% 



10% 1 5% 15% 



1. One excitatory (E) and on 
inhibitory (I) element per unit. 

2. Local connections based on 
anatomical data. 

3. Total afferent input ~ 10- 
local connections. 

4. Sigmoidal activation rule. 



The Sigmoidal Activation Rule 



dt 

dm 

dt 



r 



= A 



1 



N 



V 



-K E [w EE E t (O+w^ {t)+in lE {t)-T E +N{t)\ 



1+e 



J 



-W) 



f 



= A 



1 



A 



-Ki WeiE, (0+'% {t)-Tj +N{t)] 
where 



\l + e 



-sim 



J 



j J 

mi (0 = T}4 E k (0 + 2j% 4 (0 



Multiparameter differential equation 
■ A = rate of increase 



8 = rate of decay 



Working Memory Module (IT-PF component) 



IT 

(and other areas) 



excitatory 
inhibitory 



* e^e 



♦ e^i 



s = cue-selective 
d1 = delay 
d2 = delay+cue 
r = response 




Modulator of Attention 




Challenge for the Model 



excitatory 
inhibitory 



electrical activity 



Results and Conclusions: Visual Model 



Simulations with the Full Model 



(% CHANGE WITHIN AREAS) 

High Attention to Shape - Low Attention to Degraded Shape 



V1/V2 



+3.1% 



+5.2% +2.5% 



Experimental Results (Haxbyetai, 1995) 
+2.7% +8.1% +4.2% 



Prefrontal 



+3.5% 



+4.1% 



Conclusions 



1. Electrical activities in each region match exp. results in primates. 

2. PET activities in each region match exp. results in humans. 

3. Our hypothesis about how different frontal neuronal populations 
interact is supported. 

4. Our hypothesis about relation between integrated synaptic activity 
and PET/fMRI data is supported. 

5. Hypothesis about role of top-down processing is supported. 



Simulation of fMRI Experiments 



fMRI activity is simulated by spatial and temporal 
integration of the absolute value of the synaptic 
activity over 50 msec (which represents the time 
needed to acquire an fMRI slice). 

This time course is then convolved with a Poisson 
function representing the hemodynamic delay. 

The resulting function is then sampled every Tr sec 
(volume acquisition time) to yield the simulated fMRI 
activity during each scan series. 



(Horwitz and Tagamets, Human Brain Mapp., 1999) 



Example: Simulated fMRI data for Auditory DMS Task 



Time courses: Ai (blue) and PFC (green) 




100 




2000 



100 



6000 




2000 



50 60 
time(sec) 

S=sweep patterns R=random tones 



100 



Hemodynamic delay parameter = 6sec 



Event-Related fMRI: Dale-Buckner Study 



T1 = Duration of 1st stimulus 
T2 = Interstimulus interval 
T3 = Duration of 2nd stimulus 



T1 = 1 sec; T2 = 5 
sec; T3 = 1 sec 



19 sec 



1 

T1 


T2 


T3 


13 sec 



Stimuli: Flashing checkerboard (8Hz); Tr = 1 sec; 
hemodynamic delay parameter (X) = 6 sec. 



Event Related fMRI--Experimental Results 



Tr = 1sec; ITT = 5sec 



(Dale & Buckner, 1997) 



FIRST TRIAL 




-' 



-i 1 — i 1 — i 1 — i r— i 1 — i 1 — i — t — i f — r — i- 

1 2 3 £ 5 B 7 9 10 11 12 13 14 15 15 17 1fl 19 

TIME (SEC) 



A-> 



i — i — r 



i — r — p — r — r — i i 



~\ — r 



I 2 3 4 5 6 7 9 9 10 II 12 13 14 15 16 17 IB IB 

TIME (SEC) 



FIRST TRIAL 
/ 




~f — r — i — i — r — i — i 1 — i 1 — i r 

7 8 S 10 11 12 13 1* IS 16 ?7 10 19 
TIME (SEC) 



1 



h r 

2 3 \ 



~T 1 ■ I I 1 1 1 1 1 F I I 

6 J H 9 10 M 12 13 H '5 16 17 1ft 19 
TIMEiSEC) 



Event Related fMRI -- Simulated Results: V4 



V 



en 



4000 



3500 - 



3000 - 



2500 - 



2000 - 



1500 




4S 



"3 

c 



2000 



1500 - 



1000 - 



500 - 



0- 



(500) 




time (sec) 



time (sec) 



Tr = 1 sec; X = 6 sec; delay = 5 sec 



1st trial 
Est. 2nd trial 



Event Related fMRI - Simulated Results: Prefrontal Cortex 



5500 



I 

c 



3000 - 



2500 - 



2000 - 



1500 




1500 



3 

I 

"3 

c 

a 



20 30 



time (sec) 



Tr = 1 sec; X = 6 sec; delay = 5 sec 



1000 - 



500 - 



(500) 




time (sec) 



1st trial 
Est. 2nd trial 



Conclusions 
Simulation -- Event-Related fMRI 



Simulated and experimental results are quite similar in 
posterior brain areas. These results provide validation 
for our large-scale neural model, especially its behavior 
during small time intervals. 

Results of the simulation in anterior brain areas, 
especially the prefrontal cortex, suggest caution in 
interpreting event-related fMRI in brain areas where there 
may be substantial neural activity when stimuli are not 
present. In our simulation, the estimated second trial 
differed from the first trial due to convolving of activity 
when the stimulus was present with activity during the 
inter-stimulus interval. 



Collaborators 



Malle Tagamets 
Fatima Husain 
Antonio Ulloa 
Randy Mcintosh 
Allen Braun 
Katrin Amunts 
Karl Zilles 
Arun Bokde 
Rhonda Friedman 
Mike Glabus 
Karen Berman 
Theresa Long 



and others 

REVISION FUNDACION J.MAS