cognitive-emotional learning : an

CNP


CNP

Frontal lobe activation mediates the
activation of the amygdala during
cognitive
-
emotional learning : an
effective connectivity study


Branislava Ćurčić
-
Blake,

Marte Swart
and André Aleman


Cognitive Neuropsychiatry group,
Neuroimaging center (NIC), University
medical center Groningen (UMCG), The
Netherlands



CNP


CNP

Overview

•
Quick introduction and key points regarding DCM

•
Our emotional learning study



•
Questions and suggestions welcome at any point


CNP


CNP

Phenomenon of brain connectivity?

•

Anatomical : The connections between brain areas by means
of white matter tracts (groups of axons)


•
Functional :

Analyses of inter
-
regional effects: what are the interactions between
the elements of a given neuronal system? How functionally
specialised regions interact with each other


•
a) Functional connectivity:


the temporal correlation between
spatially remote
neurophysiological events


•
b) Effective connectivity


the influence that the
elements of a neuronal
system exert on each other

A

B

A

B


CNP


CNP

DCM

•
Neat method to establish effective connectivity
(as defined by Friston!)

•
A well
-
defined model or set of models is required

•
The fMRI data dynamics are modeled


•
Make inferences about processes that underlie
measured time series


•
Idea is to estimate parameters of a reasonably
realistic neuronal system model

such that
predicted BOLD corresponds as close as
possible to measured BOLD


From Burkhard Pleger, Functional Imaging Lab, University College London


CNP


CNP

What DCM can do and what
cannot

•
DCM can make inferences about how
much the activity in area A can induce
change of activation in area B!



•
DCM cannot make inferences about
speed of the processes, nor timing.


CNP


CNP

hemodynamic

model

effective connectivity

modulation of

connectivity

The bilinear model

Cu
z
B
u
A
z
j
j





)
(
λ

z

y

integration

Neural state equation

)
,
,
(
n
u
z
F
z



DCM

Conceptual
overview

Friston et al. 2003,
NeuroImage

u
z
u
F
C
z
z
u
u
z
F
B
z
z
z
F
A
j
j
j
























2

CNP


CNP

Important coefficients

•
A

–
Effective
connectivity

•
B

–

modulatory
effects

•
C

-

Inputs

BOLD

y

y

y

Input

u(t)

activity

z
2
(t)

activity

z
1
(t)

activity

z
3
(t)

direct inputs

c
1

b
23

a
12

neuronal

states


CNP


CNP

•
Combining the neural and
h
e
modynamic states gives the
complete

forward model
.



•
An
observation model

includes
measurement

error

e

and confounds
X

(e.g.
drift).


•
Bayesian parameter estimation



•
Result:

Gaussian a posteriori parameter
distributions
, characterised by

mean
η
θ
|y

and

covariance
C
θ
|y

and posterior covariance of
noise
C
e

.

How it works in practice:


parameter estimation

η
θ
|y

)
(
x
y


e
X
u
h
y





)
,
(
observation model

)
(
)
|
(
)
|
(



p
y
p
y
p


posterior


likelihood ∙ prior


CNP


CNP

Choosing the model

Bayes Theorem

Bayes factor


Akaike information
criterion (AIC):

Bayesian
information
criterion (BIC):


p
m
accuracy
m
y
AIC


)
(
)
|
(
Penny et al. 2004,
NeuroImage

S
N
p
m
accuracy
m
y
BIC
log
2
)
(
)
|
(


)
|
(
)
|
(
)
,
|
(
)
,
|
(
m
y
p
m
p
m
y
p
m
y
p




)
|
(
)
|
(
j
m
y
p
i
m
y
p
B
ij



Here
p
is the number of parameters and

N
s

is the number of data points


CNP


CNP

The DCM cycle
Design a study that
allows to investigate
that system
Extraction of
time series
from
SPMs
Parameter estimation
for all
DCMs
considered
Bayesian model
selection of
optimal DCM
Statistical test
on parameters
of optimal model
Hypothesis about
a neural system
Definition of
DCMs
as system
models
Data acquisition

CNP


CNP

Cognitive
-
Emotional learning study

•
What is known about emotional learning


•
Our idea


•
Our experiments


•
Results and Conclusions


CNP


CNP

Emotional learning

•
“Emotional memories constitute the core of our
personal history” (La Bar 2006)


•
Learning is enhanced or inhibited by emotions
(Phelps 2004, Richter
-
Levin 2004)

•
Emotions can

•
Enhance memory (Learning emotional words or faces;
Kensinger 2004)

•
Modulate memory (LeDeux)

•
Inhibit memory (spatial learning followed by stress
–

rats in
water maze: reviewed in Richter
-
Levin 2004)

1.
LaBar,K.S. & Cabeza,R. Cognitive neuroscience of emotional memory. Nat Rev Neurosci 7, 54
-
64 (2006).


2.
Phelps,E.A. Human emotion and memory: interactions of the amygdala and hippocampal complex. Curr. Opin. Neurobiol. 14, 198
-
202 (2004).

3.
Richter
-
Levin,G. The amygdala, the hippocampus, and emotional modulation of memory. Neuroscientist. 10, 31
-
39 (2004).

4.
Kensinger,E.A. & Corkin,S. Two routes to emotional memory: distinct neural processes for valence and arousal. Proc. Natl. Aca
d.
Sci. U. S. A 101, 3310
-
3315 (2004).

5.
LeDoux,J. The emotional brain: misterious underpinnings of emotional life. Simon & Schuster, New York (1996).


CNP


CNP

Emotional learning

•
Amygdala and Hippocampus complex are anatomically
connected (Ameral 1992; Stefanacci 1996);

•
Emotional enhancement of learning:

–
Amygdala modulates encoding and storage of Hippocampal
memories.

–
Hippocampal complex (episodic representations,
interpretations of events) can influence the amygdala response
to emotional stimuli.

•
Hippocampus
–

Amygdala effective connectivity is
modulated by positive and negative emotions during
emotional retrieval (Smith et al. 2006).

•
Amygdala modulates parahippocampal and frontal regions
during emotional memory storage (Kilpatric 2003) and
encoding item for + and
–

stimuli (Kensinger 2006) etc.


1. D. G. Amaral, J. L. Price, A. Pitkänen, S. T. Carmichael, in The Amygdala: Neurobiological aspects of emotion, memory and
mental dysfunction, J. P. Aggleton, Ed. (Wiley Liss, New York, 1992).

2. L. Stefanacci, W. A. Suzuki, D. G. Amaral, J.Comp Neurol. 375, 552
-
582 (1996).

3. E. A. Phelps, Curr.Opin.Neurobiol. 14, 198
-
202 (2004).

4. E. A. Kensinger and D. L. Schacter, J.Neurosci. 26, 2564
-
2570 (2006).

5. L. Kilpatrick and L. Cahill, Neuroimage. 20, 2091
-
2099 (2003).


CNP


CNP

•
The emotional situations influence memory on its
every stage:

–
Encoding (LeDeux 1996; Kensinger 2006)

–
Consolidation (Richter
-
Levin 2004)

–
Storage (Kilpatric 2003; Phelps 2004)

–
Retrieval (Smith 2006)

1.
LeDoux,J. The emotional brain: misterious underpinnings of emotional life. Simon & Schuster, New York (1996).

2.
E. A. Kensinger and D. L. Schacter, J.Neurosci. 26, 2564
-
2570 (2006).

3.
Richter
-
Levin,G. The amygdala, the hippocampus, and emotional modulation of memory. Neuroscientist. 10, 31
-
39 (2004).

4.
E. A. Phelps, Curr.Opin.Neurobiol. 14, 198
-
202 (2004).

5.
L. Kilpatrick and L. Cahill, Neuroimage. 20, 2091
-
2099 (2003).

6.
Smith,A.P., Stephan,K.E., Rugg,M.D. & Dolan,R.J. Task and content modulate amygdala
-
hippocampal connectivity in emotional
retrieval. Neuron 49, 631
-
638 (2006).



CNP


CNP

Brain areas involved


Functional MRI (fMRI) activation is monitored
while healthy adults encode high
-
arousing
negative words, low
-
arousing negative words
(valence only) and neutral words.


Data pooled across nine experiments
consistently show haemodynamic
changes evoked by conditioned fear
stimuli in the amygdala and subjacent
periamygdaloid cortex (coronal
sections, left), and the thalamus and
anterior cingulate/dorsomedial
prefrontal cortex (ACC/DMPFC, mid
-
sagittal section, right).

LaBar,K.S. & Cabeza,R. Cognitive neuroscience of emotional memory. Nat Rev Neurosci 7, 54
-
64 (2006).


CNP


CNP

MFcortex and IFcortex

•
Emotional memory studies show also involvement
of MFC and IFC

•
MFC is sensitive to tasks involving emotions,
mental state attribution (1), monitoring for and
detecting errors (2), and mentalizing (3).

•
IFC is engaged in emotion regulation, processing
semantic aspects of face recognition, and
language tasks. The left IFG selects the task
-
relevant information (emotional connotation as
target information from specific competing
semantic alternatives; 4).


1.
Olsson,A. & Ochsner,K.N. The role of social cognition in emotion. Trends Cogn Sci. 12, 65
-
71 (2008).

2.
Summerfield,C. et al. Predictive Codes for Forthcoming Perception in the Frontal Cortex. Science 314, 1311
-
1314 (2006).

3.
Amodio,D.M. & Frith,C.D. Meeting of minds: the medial frontal cortex and social cognition. Nat Rev Neurosci 7, 268
-
277 (2006).

4.
Ethofer,T. et al. Cerebral pathways in processing of affective prosody: A dynamic causal modeling study. NeuroImage 30, 580
-
587
(2006)


CNP


CNP

Model of Amygdala involvement in
Emotional Learning

•
Potential mechanisms by which the amygdala mediates the
influence of emotional arousal on memory
.


LaBar,K.S. & Cabeza,R. Cognitive neuroscience of emotional memory. Nat Rev Neurosci 7, 54
-
64 (2006).


CNP


CNP

Our aim

•
how the emotions and cognition interact during
cognitive emotional learning ?


•
whether the emotions revealed by activation of
the amygdala modulate the way in which the
cognition works during an associative emotional
learning task that engages
HIGHER COGNITIVE
PROCESSES

during the learning of emotional
stimuli.

•
We incorporate both positive and negative
emotional stimuli in order to see whether these
circles differ and if so, how.


CNP


CNP

Starting point

•
Data obtained from experiments by Marte Swart

•
Students: 20 LOW score on BVAQ (Bermond
-
Vorst Alexithymia Questionnaire).

•
An emotional picture
-
word associate learning task
(ALT)

•
Thus cognitive emotional processing



CNP


CNP

The task

Roomijs

(= ice
-
cream in English)
)

Do

picture and word fit?

Memorize

2
-
8sec

3sec

Task ALT

•
An emotional picture (International Affective
Picture System) and a word were displayed for 3
seconds.

•
2
-
8 seconds to
decide

if the word and picture
fitted together AND to
remember

them


CNP


CNP

Results

•
bilateral amygdala (AMY),

•
inferior frontal gyrus (IFG),

•
medial frontal gyrus (MFG), and

•
fusiform gyrus (FG) during the ALT.

RFX analysis


ALT emotional > neutral


for low
-
alexithymia subjects


(p<0.005, T>2.92, unc.).


Crosshair
[12,
-
16,
-
14], MNA.


FGR

AmyR

AmyR

IFGR

MFGR


CNP


CNP

The DCM ROI selection

Fig. 3 Contrast as it is
used to define VOI’s: ALT
emotional >fixation point
(random effects t
-
test) for
20 subjects. The IFG,
MFG and Amy are circled
for illustration (p<0.001,
T>3.3, unc.).

Crosshair [
-
22,
-
4,
-
16],
MNI.


AmyL

AmyL

AmyL

IFGL

IFGL

MFGL


CNP


CNP

The maximum activation per ROI



BA

x

y

z

Z

AMY L



-
22

-
4

-
16

4.45

AMY R



22

-
4

-
18

4.8

IFG L

45

-
56

22

14

7.01

IFG R

45

56

28

18

4.63

MFG L

10

-
6

8

50

7.28

MFG R

10

6

8

50

6.2


CNP


CNP

The creation of VOI’s

•
The VOI’s for each subject !

•
created by choosing the closest supra
-
threshold
(
p

< 0.05
) voxel

•
within the Maximum Probability Maps (of the
Anatomical Toolbox in SPM5)

•
Belongs to the region (visual inspection)

•
Sphere of 4 mm drawn around

•
10
-
33 voxels

•
Time series extracted

•
1
st

Principal Component (PA)


CNP


CNP

Checking for input

Fig 4. Full DCM models with different areas of input for the selection of input area(s).
Input is illustrated by black arrows, and effective connectivity by grey arrows.

Model M

MF

IF

AMY

Model I

MF

IF

AMY

Model A

MF

IF

AMY

Model MI

MF

IF

AMY

Model AI

MF

IF

AMY

Model MA

MF

IF

AMY

Model


ABf

PER

MI/M L

4.6*10
11

8/4

MI/M R

26.8

6/2

MI/I L

2.6*10
12

7/6

MI/M R

3.06

7/5

MI/A L

3.8*10
11

8/4

MI/A R

14.4

8/2

MI/MA L

20.4

10/7

MI/MA R

74

13/3

MI/AI L

16.2

11/4

MI/AI R

135

14/2


CNP


CNP

Choosing the best connectivity mod

Fig 5.Illustration of models of effective connectivity during an ALT. Input consisting of
positive, negative and neutral conditions goes parallel to the IFG and MFG. In Model #1
the IF and MF communicate directly and with the Amy as opposed to #2 and #3 where
the IF and MF communicate through the Amy. Models #4,5 and 6 are variations of model
#1. The winning model #1 was also compared to the full MI model. The results are
presented in Table 3.

Model #1

IF

MF

AMY

Model #2

IF

MF

AMY

Model #3

IF

MF

AMY

Model #4

IF

MF

AMY

Model #6

IF

MF

AMY

Model #5

IF

MF

AMY


CNP


CNP

The resulting connectivities and
modulatory effects

Mean

SD

t

Sig.

MFG to Amy

.16

.12

5.9

<.00

MFG to IFG

.27

.33

3.4

.003

IFG to MFG

.21

.35

2.6

.02

IFG to Amy

.16

,15

4.6

<.00

Pos MFG to Amy

-
.01

.08

-
.7

.5

Pos MFG to IFG

.01

.08

0.4

.6

Pos IFG to MFG

.01

.06

1

.3

Pos IFG to Amy

-
.015

.07

-
0.9

0.4

Neg MFG to Amy

-
.03

.08

-
1.6

.1

Neg MFG to IFG

.04

.07

2.52

.02

Neg IFG to MFG

.04

.08

2.45

.03

Neg IFG to Amy

-
.03

.08

-
1.5

.2

neu MFG to Amy

-
.02

.097

-
.9

.35

neu MFG to IFG

.04

.07

.5

.6

neu IFG to MFG

.03

.09

1.4

.2

neu IFG to Amy

-
.003

.1

-
.1

.9

Table 4

b. Right

Mean

SD

t

Sig.

MFG to Amy

.12

.14

3.8

.001

MFG to IFG

.2

.3

3.0

.008

IFG to MFG

.2

.3

2.5

.02

IFG to Amy

.14

,15

4.5

<.00

Pos MFG to Amy

.01

.05

-
.8

.4

Pos MFG to IFG

.02

.06

1.6

.1

Pos IFG to MFG

.03

.07

2.2

.04

Pos IFG to Amy

-
.03

.05

-
2.2

0.04

Neg MFG to Amy

-
.01

.07

-
.5

.6

Neg MFG to IFG

.05

.08

2.97

.008

Neg IFG to MFG

.02

.03

2.99

.008

Neg IFG to Amy

.01

.09

-
0.7

.5

neu MFG to Amy

-
.03

.07

-
1.9

.07

neu MFG to IFG

.01

.07

.8

.4

neu IFG to MFG

.04

.08

2.3

.03

neu IFG to Amy

-
0.04

.07

-
.45

.02

Table 4 a. Left


CNP


CNP

The resulting connectivities and
modulatory effects

Fig 6. Modulatory effects of the best DCM model. Increasing effect (red bold arrows) and
decreasing effect (blue dashed arrow) are presented with the % of influence on the effective
connectivity and the significance level (in brackets).

Amy
N(15%;0.02)
N(20%;0.03)
IF
MF
N(27%;0.008)
P(18%;0.05),
N(13%;0.008),
n(22;0.03)
P(-17%,0.05),
n(-26%;0.02)
Amy
IF
MF

CNP


CNP

Lateralization yes or no?

•
t

–

test between modulatory effects for left and
right hemisphere showed NO significant
difference between mean values of modulatory
effects for each pair.


•
Thus, we can not claim that there is lateralization.


CNP


CNP

Conclusions

•
The area involved in basic emotional learning (the
amygdala) does not affect the change in activity
of the cognitive areas (the IFG and MFG).

•
The subjects appear to pay more attention to the
context and evaluation of the given stimuli, and
these processes were not affected by emotions.

•
In our case it seems that the subjects
concentrated on the task and suppressed their
emotions to some extent


CNP


CNP

Main conclusion

•
In conclusion, it is evident that complex emotional
learning is led by a “top
-
down” process from the
frontal areas
-

the MFG and IFG
-

to the amygdala.


CNP


CNP

Pitfall

•
There are gender differences (Cahil et al
2001;2002):

–
Correlations between Left Amygdala
–

emotional
memory enhancements for Females

–
Correlations between Right Amygdala
–

emotional
memory enhancements for Males


•
We found no significant gender differences due to
low statistical power for such a comparison






CNP


CNP

Still

•
We have demonstrated that complex emotional
learning is led by top
-
down processes from the
frontal lobe toward the amygdala.

•
This type of learning is more complicated than
conditioned fear therefore the learning circuit is
more complex.

•
The top
-
down processes demonstrate that the
cognition here is “emotion free”.

•
The amygdala might still play a role in the
modulation of learning material delivered to the
memory areas. (it does! data not shown here ;
-
D)