THE EFFECT OF SAD MOOD ON PAIN PERCEPTION: IMAGING FUNCTIONAL BRAIN
NETWORKS
By
Lan Yang

A DISSERTATION
Submitted to
Michigan State University
In partial fulfillment of requirements
For the degree of
DOCTOR OF PHILOSOPHY
Neuroscience
2012

ABSTRACT
THE EFFECT OF SAD MOOD ON PAIN PERCEPTION: IMAGING FUNCTIONAL BRAIN
NETWORKS
By
Lan Yang
People with chronic pain are often afflicted with psychiatric depressive disorders. Even the
commonplace, everday experience of pain can be modified by ordinary emotion. However, the
neural mechanisms underlying these phenomena are still unclear. Investigating the effects of sad
mood on pain perception at the behavioral and the neural levels helps to understand the comorbidity of depression and pain problems. This dissertation focuses on: (1) examining the
effects of sad mood on perception of painful stimuli, and (2) identifying the functional brain
networks that could provide the neural substrate for pain perception altered by sadness.
In Chapter 2, a task-evoked functional magnetic resonance imaging (fMRI) method has been
used to investigate which brain areas are activated by painful and emotional stimuli. The
hypotheses are that the subjective perception of pain intensity would be greater during sadness
compared to neutral mood, and significantly increased neural activation would be observed in the
primary cortical areas for pain perception. To test these hypotheses, the fMRI study has recruited
sixteen healthy female subjects. The experimental trials include the delivery of painful electrical
shocks concurrent with or without sad mood induced by the presentation of pictures from the
International Affective Picture System (IAPS). The study identifies widely-distributed cortical
and subcortical areas involved in the processing of pain and the altered pain perception by
sadness. The cortical areas for pain processing are mainly located in the primary/secondary
somatosensory, insular and cingulate cortices. Specifically, the posterior insular and the adjacent

secondary somatosensory cortices (pIn/SII) have significantly increased neural activation, when
sadness intensifies subjective perception of pain.
In Chapter 3, using the resting-state fMRI and path analysis, it further examines the
functional networks of pain processing and altered pain perception by sadness. The pain-relevant
brain areas, such as primary somatosensory cortex, anterior cingulate cortex and thalamus, are
intrinsically correlated with pIn/SII during rest. The path coefficients in the pain processing
functional brain network increase during painful stimuli. The path analysis has also revealed a
functional network in which the pIn/SII connects with emotion-relevant areas, such as the
subgenual cingulate cortex, anterior insula and amygdala. These findings provided evidence of
specific neural pathways that may be relevant to understanding the comorbidity of depression
and pain problems.

DEDICATION

To Tao and Vesna

iv

ACKNOWLEDGEMENTS

The writing of this dissertation has been one of the most significant academic challenges I
have ever had to face. Without the help, support and guidance of the following people over years
in Michigan State University, this study would not have been completed. It is to them that I owe
my deepest gratitude.
First of all, I am extremely lucky to have Dr. Laura Symonds as my advisor, who committed
years in training me to be a researcher, provided me an excellent atmosphere for doing research
and financially supported my research. Her wisdom, knowledge and commitment to the highest
standards inspired and motivated me. She offered valuable suggestions in guiding my research
directions. She read and edited my manuscripts and dissertation drafts, as well as corrected my
grammar errors without complaints. It is an honor to be her student and to work with her. I am
also grateful for her influence on me to face challenges in school and in life confidently,
persistently and optimistically. What I have learned from her would be my lifelong treasure
carried with me after graduation.
I would like to thank Dr. David Zhu, a member of my Ph.D committee, for guiding me in
designing experiments, collecting resting-state fMRI data, and patiently correcting my writing. I
also appreciate the help from other members of my Ph.D committee: Dr. Gregory Fink and Dr.
Sharleen Sakai. They gave the knowledge and insights in guiding my research and helping me to
develop my research background in statistics, psychology, physiology and neuroanatomy. They
patiently guided and encouraged me in the process of this dissertation.
I specially thank my lab manager, Lynn Mande, for everything. Without her support, I could
not have accomplished anything. I would like to thank my fellow lab-mates, Ramin Rajaee and
James Murphy, for many useful discussion, help and friendship.

v

I am grateful to the neuroscience program and the department of radiology of Michigan State
University for financially supporting my doctoral training and providing experimental facility for
my dissertation study. I thank the undergraduate students of the Michigan State University for
participating in my experiments. I also thank all the professors at Michigan State University
whom I have taken courses from and interacted with. I also thank all the friends I made in the
neuroscience program. Thank you all for making my graduate-student life cheerful.
Last, I would like to thank my family. They are always supporting me spiritually,
emotionally and psychologically. I would like to thank my husband, Tao Wu, for encouraging,
supporting and loving me unconditionally during the past years. My parents, ShulinYang and
Changfen Zheng, receive my deepest gratitude for their understanding and supporting during my
undergraduate and graduate studies. I will never forget how difficult it was for them to accept my
decision to leave China for my dream. I thank them for their faith in me and respecting every
decision I made for my life. I also thank Tao’s parents for unending encouragement and support.
I wish I could enumerate all the people I need to give thanks, but I am sure I missed many of
them on this short acknowledgement. I am grateful for them and wish I could pass their kindness
to other people in the future.

vi

TABLE OF CONTENTS

LIST OF TABLES ……………………………………………………………………………….ix
LIST OF FIGURES ………………………………………………………………………………x
Chapter 1. Introduction…………………………………………………………………………1
1.1 Pain and its Co-morbidity with Depression………………..…………………..…….………..1
1.2 The Phenomenon of Pain…………………………………………..……………..…………...2
1.2.1 Multi-dimensional Characteristics of Pain………………………………………………...2
1.2.1.1 The Sensory Dimension of Pain …………………..…………………….…………....3
1.2.1.2 The Affective Dimension of Pain…………………………………………..…………3
1.3 The Neural Mechanisms of Pain Processing………………………………………………….4
1.3.1 Primary Ascending Pathways of Pain Transmission………………….…………………..5
1.3.1.1 Spinothalamic and Spinomesencephalic Pathways ….....…………………………….6
1.3.1.2 Parallel Systems Associated with Sensory and Affective Dimensions of Pain ......…..8
1.3.2 The Neuromatrix Theory of Pain……...…………………………………………………..9
1.3.3 Brain Regions of Pain Processing ……………………………………………………….11
1.3.3.1 Brain Regions Associated with the Sensory Dimension of Pain……………….....…11
1.3.3.2 Brain Regions Associated with the Affective Dimension of Pain….....................…..12
1.3.3.3 Supraspinal Modulation of Pain Transmission………………………………...…….14
1.4 The Neural Mechanisms of Emotions…………………………………………………….….15
1.4.1 Neuroanatomical Theories of Emotions…………...…………………………………….15
1.4.2 Brain Regions Associated with Emotions ……….………………………………………17
1.5 The Interaction of Pain and Emotions ...………………………………………………….....19
1.5.1 Emotions Influence the Perception of Pain …….……………………………….……….19
1.5.2 Neural Substrate correlated with the interaction between Pain and Emotions……….….20
1.6 Background to the Methodology……….……………………………………………..……..22
1.6.1 Task-driven fMRI and Data Analysis……………………………………………….…...23
1.6.2 Resting-state fMRI and Data Analysis…………………………………………………..26
1.6.3 Functional Connectivity of Pain Network…………………………………………….…29
1.6.3.1 Path Analysis……………………………...………………..…………………..……29
1.6.3.2 Anatomical Connections of Pain and Emotion Networks………………….….…….30
1.7 Research Questions of this Dissertation ………………..………………………………...…32
Chapter 2. An fMRI Study of Pain Perception Modulated by Sad Mood………………….34
2.1 Introduction ………………………………………………………………..……………..….34
2.2 Methods…………………………………………..…………………………..………………35
2.2.1 Subjects…………………………………..…………………………………………..…..35
2.2.2 Experimental Procedure ……………………………..………………………..…………36
2.2.3 Image Acquisition ……………………………………..………………………..……….38
2.2.4 Image Analysis ………………………………………………………………..…..……..39
2.3 Results…………………………………………………………………………………..……41
2.3.1 Behavioral Data ………………..……………………………….…………………….…41

vii

2.3.2 Neuroimaging Data ……………………………………………………..……………….43
2.3.2.1 Brain Regions Activated During Experimental Tasks…………………………….…43
2.3.2.2 The Interaction of Sadness and Pain …………………...…………………….……...49
2.4 Discussion …………………………………………………………………………..……….51
2.4.1 Increased Activation in Posterior Insular and Somatosensory Cortices...……………….52
2.4.2 The Role of Anterior Cingulate Cortex…………………………………………….……53
2.4.3 Sadness and the Decending Pain Pathway……………………………………………….55
2.4.4 Conclusion ………………………………………………………………………………58
Chapter 3. Functional Connectivity of Brain Networks for Pain Processing and for
Interaction of Pain and Sadness ………………………………………………………....……59
3.1 Introduction ……………………………………………………………………………….…59
3.2 Methods……………………………………………………………………………..……..…61
3.2.1 Subjects……………………………………………………………………………….….61
3.2.2 Image Acquisition ……………………………………………………………..……..….61
3.2.3 fMRI Data Preprocessing and Analysis………………………………………………….62
3.2.4 Simple Correlation analysis and group ICA analysis……………………...............…….64
3.2.5 Functional Connectivity Models …………………………………………………...……66
3.3 Results………………………………………………………..………………………………70
3.3.1 Resting-state Maps……………………………………………...………………………..70
3.3.2 Connectivity Model of Pain Network……………………………...…………………….76
3.3.3 Connectivity Model for Interaction of Pain and Sadness……...………………..……….76
3.4 Discussion………………...………………………………………………………………….80
3.4.1 Functional Network for Pain Perception …………………………………..……………80
3.4.2 Functional Network for Interaction of Pain and Sadness ……………………….........…82
3.4.3 Conclusion ………………………………………………………………………………84
Chapter 4. General Discussion ....……....…………………………………………………….. 86
4.1 Brain Regions Encoding both Pain and Emotions……………………….…..………………86
4.2 Secondary Somatosensory and Posterior Insular Cortices…………………...………………88
4.3 Functional Connectivity Analysis of Pain-Emotion Interaction ..…………………..……….91
4.4 Conclusions…………………………………………………..……………...……………….93
4.5 Limitations……………………………..…………………………………………………….94
4.6 Future Directions…………………………………………………………………………….94
APPENDICES………………………..…………………………………………………………96
APPENDIX A………………………..…………………………………………….……………97
APPENDIX B…………………………………………………………………….……………104
REFERENCES ……………..……………………………..………………………………..…113

viii

LIST OF TABLES

Table 2.1 Brain regions activated during three experimental trials……………………………...48
Table 2.2 Brain activation as results of contrast analyses of (Sad and Pain – Sad Only – Pain
Only) and (Sad and Pain – Sad Only)……………………………………………………………51
Table 3.1 Regions of Interest (ROIs)…………………………………………………………….64
Table 3.2 Input data for analyzing the functional connectivity of the pain processing network
(Figure 3.2) during rest. Pearson correlations between the ROIs (** p < 0.01)…………………77
Table 3.3 Input data for analyzing the functional connectivity of the pain processing network
(Figure 3.2) during painful stimulation. Pearson correlations between the ROIs (** p < 0.01)...78
Table 3.4(A) Input data for analyzing functional connectivity of Figure 3.3 when painful stimuli
were administered during sad mood. Pearson correlations between the ROIs (** p < 0.01)……78
Table 3.4(B) Input data for analyzing functional connectivity of Figure 3.3 when painful stimuli
were administered during sad mood. Pearson correlations between the ROIs (** p < 0.01)……78

ix

LIST OF FIGURES

Figure 1.1 Two major ascending pathways for nociceptive processing: the spinothalamic
pathway is shown on the left, and the spinomesencephalic pathway is shown on the right………6
Figure 1.2 Flow charts of a task-driven fMRI experimental procedure…………………………26
Figure 1.3 Cerebral cortical pathways involved in the pain processing…………………………31
Figure 2.1 Block design of fMRI trials: Pain Only, Sad Only, and Sad and Pain. Dotted black
line in Trial 1 and Trial 3 indicates delivery of 6 pulses of electrical shock, each presented for 3
seconds with 2 seconds between pulses………………………………………………………….38
Figure 2.2 Mean ratings of sadness before and after each experimental task. The ‘*’ indicates
significant difference in the paired t-test (P < 0.0001)…………………………………………..42
Figure 2.3 Mean ratings of pain intensity before and after each experimental trial. The ‘*’
indicates that pain was rated significantly more intense for the Sad and Pain trial than Pain Only
in the paired t-test (p < 0.001)……………………………………………………………………42
Figure 2.4 Brain activation during the Pain Only condition. Images in all three rows left to right:
axial, coronal and parasagittal. Left hemisphere shown on left in axial and coronal views. Green
crosshairs centered on region of interest. Row A: left primary somatosensory cortex (SI); Row B:
bilateral posterior insular/secondary somatosensory cortices (pIn/SII); Row C: left anterior
cingulate gyrus (BA 32). Alpha value, corrected for whole brain, is 0.001. (For interpretation of
the references to color in this and all other figures, the reader is referred to the electronic version
of this dissertation.)………………………………………………………………………………44
Figure 2.5 Brain activation during the Sad Only condition. Images in all three rows left to right:
axial, coronal and parasagittal. Left hemisphere shown on left in axial and coronal views. Green
crosshairs centered on region of interest. Row A: bilateral medial prefrontal gyrus (BA 9); Row
B: right inferior frontal gyrus (BA 45/44); Row C: bilateral amygdala/hippocampal complex.
Circles in the axial image indicate bilateral occipital/fusiform gyri. Alpha value, corrected for the
whole brain, is 0.001……………………………………………………………………………..45
Figure 2.6 Brain activation during the Sad and Pain condition. All images are in the coronal
plane. Green crosshairs centered on region of interest. Left hemisphere shown on left. A:
bilateral posterior insular/secondary somatosensory cortices (pIn/SII); B: left anterior cingulate
gyrus (BA 32). C: right medial prefrontal cortex; D: bilateral amygdala/hippocampal complex.
Alpha value, corrected for the whole brain, is 0.001…………………………………………….46
Figure 2.7 Brain activation during the Sad and Pain condition. Images in all three rows left to
right: axial, coronal and parasagittal. Left hemisphere shown on left in axial and coronal views.
Green crosshairs centered on region of interest. Row A: left periaqueductal grey (PAG); Row B:

x

left thalamus; Row C: right anterior insular cortex. Alpha value, corrected for the whole brain, is
0.001…………………………………………………………………………………………….. 47
Figure 2.8 Brain activation as a result of the contrast analysis (Sad and Pain – Sad Only – Pain
Only). All images are in the axial plane. Green crosshairs centered on region of interest. Left
hemisphere shown on left. A: bilateral subgenual anterior cingulate cortex; B: left posterior
insular/secondary somatosensory cortices (pIn/SII); C: left periaqueductal grey (PAG); D:
bilateral amygdala/hippocampal complex. Alpha value, corrected for the whole brain, is
0.005……………………………………………………………………………………………...50
Figure 2.9 Brain activation as a result of the contrast analysis of (Sad and Pain – Sad Only).
Images are in the parasaggital plane. Green crosshairs centered on region of interest. A: left
anterior cingulate cortex (BA 32); B: left posterior insular/secondary somatosensory cortices
(pIn/SII); C: right pIn/SII. Alpha value, corrected for the whole brain, is 0.001………………..50
Figure 3.1 Schematic of ascending pathways, subcoritical structures and cortical structures
involved in processing pain (Price, 2000). It is as same as Figure 1.3 in Chapter 1…………….63
Figure 3.2 A hypothetical model used to investigate the functional connectivity in the supraspinal
network involved in pain processing. SI, primary somatosensory cortex; pIn/SII, posterior
insular/secondary somatosensory cortex; PCC, posterior cingulate cortex (BA 30); ACC, anterior
cingulate cortex (BA 24); Tha, ventral lateral posterior nucleus of thalamus; PAG,
periaqueductal gyrus……………………………………………………………………………..69
Figure 3.3 A hypothetical model used to investigate the interaction of pain and emotion. This
model combines the brain regions involved in both pain and emotional processing. SI, pIn/SII,
PPC, Tha, and PAG are brain regions mainly associated with pain processing. IFG, mPFC, and
OCP are involved in processing emotional information. The ACC, aIn and AMY/Hip ROIs (bold)
are the probable key nodes modulating the pain experience by emotional states……………….70
Figure 3.4 Coronal views in standard Talairach stereotactic space of whole-brain interregional
functional connectivity during the resting state at a group-level correlation analysis. A predefined seed ROI was placed in the posterior insular/secondary somatosensory cortex, P < 0.001,
correlation coefficient r > 0.2. Positive correlations of spontaneous fluctuations between the
brain voxels and the ROI voxels are shown in yellow, orange and red. These regions included: 1,
inferior frontal junction/anterior insular cortex; 2, anterior cingulate gyrus (BA 24); 3, inferior
frontal and superior temporal gyri, anterior/middle insular cortex; 4, posterior insular/secondary
somatosensory cortices; 5. primary somatosensory cortex/parietal lobule; 6. thalamus; 7.
secondary somatosensory cortex/parietal lobule………………………………………………...72
Figure 3.5 Coronal views in standard Talairach stereotactic space of whole-brain inter-regional
functional connectivity during the resting state at a group-level correlation analysis. A predefined seed ROI was placed in the anterior cingulate cortex, P < 0.001, correlation coefficient r
> 0.2. Positive correlations of spontaneous fluctuations between the brain voxels and the ROI
voxels are shown in yellow, orange and red. These regions included: 1. insular cortex; 2.
thalamus………………………………………………………………………………………….73

xi

Figure 3.6 Sagittal views in standard Talairach stereotactic space of whole-brain interegional
functional connectivity during the resting state at a group-level correlation analysis. A predefined seed ROI was placed in the medial prefrontal cortex, P < 0.001, correlation coefficient r
> 0.2. Positive correlations of spontaneous fluctuations between the brain voxels and the ROI
voxels are shown in yellow, orange and red. These regions included: 1. anterior insular cortex; 2.
prefrontal and anterior cingulate cortices; 3. posterior cingulate cortex………………………...74
th

th

Figure 3.7 A. The 10 independent component map and B. the 17 independent component
map (P<0.001, z = 2.06). The temporal patterns of regions in yellow/orange have positive
component weights…………………….………………………………………………………...75
Figure 3.8 Functional connectivity for the pain network during rest. Standardized path
coefficients, computed from data in Table 3.2, represent the combined effect of all connections
included in the path analysis. The final model is modified based on the hypothetical model in
Figure 3.2………………………………………………………………………………………...79
Figure 3.9 Functional connectivity for the pain network during acute painful stimulation.
Standardized path coefficients, computed from data in Table 3.3, represent the combined effect
of all connections included in the path analysis. Bold arrows indicate increased path coefficients
of functional connectivity during painful stimulation compared to resting state……………......79
Figure 3.10 Functional connectivity for interaction between painful and sad conditions.
Standardized path coefficients, computed from data in Table 3.4, represent the combined effect
of all connections included in the path analysis. The ROIs primarily associated with the
processing of pain perception, including SI, pIn/SII, PPC and Tha, are indicated by the dashed
double line boxes. The ROIs primarily associated with the processing of sad pictures, such as
ACC, sACC, aIn, PCC and AMY/Hip, are indicated by the dashed line boxes. PAG is a node that
receives inputs from emotional brain regions and is in a position to directly influence activation
in the somotosensory cortex. This final model includes statistically significant paths and is
modified based on the hypothetical model in Figure 3.3………………………………………...80

xii

Chapter 1
Introduction
1.1 Pain and its Co-morbidity with Depression
In the United States, pain is one of the largest medical health problems and results in high
societal costs. For example, low back pain alone anually costs $90 billion dollars in physician
services, hospital costs and pharmacy expenditures, as well $7 billion dollars indirect costs in
sick leave, early retirement, and lost household productivity (Dagenais et al., 2008).
Depression is frequently co-morbid with chronic pain. For example, patients with complaints
of persistent pain are four times more likely to be diagnosed with an anxious or depressive
disorder than those who are pain-free (Gureje et al., 2001). In primary care practice, up to 80%
of depressed patients present with physical symptoms that include headache, abdominal pain and
musculoskeletal pain (Stahl, 2002). Furtermore, there is evidence that both depression and
chronic pain increase the risk for a subsequent first onset of the other. For example, in a cohort
study examining the relationship between migraine and major depression during a 2-year period,
researchers found that the risk of first-onset migraine in persons with pre-existing major
depression was threefold higher than in persons with no history of major depression. In addition,
the risk of first-onset major depression in persons with pre-existing migraine was more than
fivefold higher than in persons with no history of headaches (Breslau et al., 2003).
The neural mechanisms underlying the co-morbidity of pain and depression remain unclear.
Cognitive-behavioral training techniques and antidepressant medicine (e.g. selective serotonin
reuptake inhibitors (SSRIs)) have been used to effectively treat individuals with either chronic
pain or depression (Blanchard et al., 1985; Gallagher et al., 1999). Examples such as this suggest
that pain and depression may share some common neural pathways.

1

1.2 The Phenomenon of Pain
Pain sensation, including pricking, burning, aching, stinging and soreness, is a sub-modality
of somatic sensation and serves an important protective function to organisms by helping them to
avoid injury and permanent damage. Unlike other somatic sensations, pain has a primitive
quality intimately associated with emotional experiences. Moreover, the intensity of perceived
pain can be influenced by emotional context so that the very same noxious stimulus may be
perceived differently under different conditions and for different individuals. In this way, pain is
a subjective perception. According to the International Association for the Study of Pain (IASP),
pain is an “unpleasant sensory and emotional experience associated with actual or potential
tissue damage, or described in terms of such damage” (Merskey and Bogduk 1994; p. 210). To
better describe the necessary and sufficient associations among experiences of unpleasantness,
sensation and tissue damage, a revised definition of pain was proposed by Price (1999) who
suggested that pain is “a somatic perception containing (1) bodily sensation with qualities like
those reported during tissue-damaging stimulation, (2) an experienced threat associated with this
sensation, and (3) a feeling of unpleasantness or other negative emotion based on this
experienced threat.” In other words, a painful experience consists of both bodily sensation and
negative feelings. Further, it is not necessary to objectively demonstrate actual or potential tissue
damage: pain can occur without peripheral nociception. If a person describes his/her experience
as pain, it is defined as pain regardless of actual tissue damage. This definition of pain highlights
the importance of investigating the relationship between pain sensation and emotions.
1.2.1 Multi-dimensional Characteristics of Pain
In 1968, Melzack and Casey proposed a conceptual model of pain that explicitly formulated
how the various dimensions of pain could be transmitted in the nervous system. Their model of

2

pain included sensory, affective and cognitive dimensions (Melzack and Casey, 1968).
According to this model, the sensory dimension consists of a bottom-up sensory-discriminative
system that provides information about the location, intensity and duration of painful stimuli, as
well as the type of pain (e.g. burning or stinging). The affective dimension consists of a
behavioral motivational system that is responsive to the aversive drive and affective
characteristics of pain. This behavioral motivational system provides information about the
unpleasantness of the painful sensation and drives the motivation to escape. The cognitive
dimension consists of a cognitive control system that regulates the intensity of activation in the
discriminative and motivational systems, shapes attitudes and beliefs towards pain, and drives
decisions and strategies to modulate it. The interaction of sensory, affective and cognitive
domains of pain thus determines our perception of a painful stimulus. The model is generally
accepted in the field of pain research and its application will be described in the following
sections, focusing on the sensory and affective dimensions of pain.
1.2.1.1 The Sensory Dimension of Pain
The sensory qualities of pain are diverse. When peripheral tissues are intensively stimulated,
sensory receptors (nociceptors) are activated. The ensuing nociceptive sensations can be
described according to their intensity, location and duration. For example, some types of pain are
experienced as pressure, and described in such words as “pinching”, “cramping” or “pricking”.
In terms of duration, pain can be acute or tonic. For example, if one touches a hot stove and
immediately removes his or her hand, the pain is sharp and hot. If one’s hand is not removed
immediately, resulting in deep tissue injury, persistent pain can arise from accompanying
inflammation.
1.2.1.2 The Affective Dimension of Pain

3

Pain is an unpleasant sensory and emotional experience. Nociception only contributes part to
the total pain experience, mainly because of the affective qualities of pain. For example, if a
person suddenly receives an electrical shock, the unpleasant sensation can also be accompanied
by visual attention to the electrical plug, autonomic responses, memories of past experiences, and
emotions such as annoyance. All these responses result in the experience of, for instance, intense
threat and subsequent fear. Importantly, it is possible to experience pain without nociceptive
inputs. For example, we can experience another person’s pain. This empathy for pain primarily
involves the affective pain dimension (Singer et al., 2004).
1.3 The Neural Mechanisms of Pain Processing
In 1664, the reflex theory of Descartes proposed a direct-line pain pathway from pain
receptors in the body to a pain center in the brain (see reviews in Deleo, 2006). This theory was
the beginning of the modern doctrine of reflexes and the first documented attempt to understand
pain transmission. Descartes’s view of the human pain system directed pain treatment for more
than 330 years. Studies during this same period investigated physiological mechanisms at every
level of the direct pathway, from receptors to the cerebral cortex. The theory implied that the
simple cutting of this pathway should alleviate pain. However, clinical evidence has shown that
this manipulation does not always alleviate pain. Damage to nerves, instead of alleviating pain,
can cause serious neuropathic pain at some point. Amputees often have the illusion that their
missing limbs are still present and most experience phantom pain: a tingling or burning sensation
in their missing limbs (Melzack, 1989). This phenomenon suggests that neural mechanisms of
the pain experience are more complex than those described by the direct-line pain pathway.
A major reinterpretation of how pain is transmitted to the nervous system and how the neural
architecture controls pain intensity came in 1965 with Melzack’s and Wall’s gate control theory

4

of pain (Melzack and Wall, 1965). The gate control theory postulates that the perception of pain
is not simply the direct result of the activation of nociceptive receptors, but involves the
interaction of different sets of neurons in the spinal cord. The idea is that pain results from the
balance of activity of nociceptive and non-nociceptive afferents in the spinal cord. Neurons in
lamina V and I receive convergent excitatory inputs from both nociceptive Aδ/C fibers and nonnociceptive Aβ fibers. It was proposed that Aβ fibers activated inhibitory interneurons in lamina
II, which could then inhibit the firing of neurons in lamina V. In contrast, Aδ and C fibers inhibit
interneurons in lamina II. In the spinal cord, a gate to the central transmission of noxious input is
therefore “closed” by non-nociceptive afferents and is “opened” by nociceptive afferents. Neural
gates in the spinal cord can be opened or closed by sensory stimuli ascending from the body.
1.3.1 Primary Ascending Pathways of Nociceptive Pain Transmission
To generate nociceptive pain sensation, information about tissue damage is transferred from
the spinal cord to the brain through five major pathways: the spinothalamic, spinomesencephalic,
spinoreticular, cervicothalamic, and spinohypothalamic pathways (Price, 1999). Within the brain,
the destinations of the different pain-related ascending pathways have bearing on differential
processing of separate pain components. The spinothalamic and spinamesencephalic pathways
(Figure 1.1) are the major pathways important for sensory and affective dimensions of pain that
are described in the following sections. The spinoreticular pathway originates from the spinal
cord and terminates in both the reticular formation and the thalamus. This pathway is likely
related to pain-related arousal and motor responses (Willis and Westlund, 1997).

The

cervicothalamic pathway arises from the lateral cervical nucleus and ascends to the
ventroposterior lateral nuclei of the thalamus. Studies in both humans and rats demonstrate that
this pathway signals pelvic visceral pain (Hirshberg et al., 1996). Finally, the spinohypothalamic

5

pathway projects directly to the hypothalamus and the supraspinal autonomic control centers,
and is considered to contribute to autonomic and endocrine responses to painful stimuli (Bernard
et al., 1996).

To associate cortex

Somatosensory
Cortex

Thalamus:
Central
Lateral
nucleus
Ventral
posterior
lateral
nucleus

Reticular
formation
Inferior
colliculus

Superior
colliculus Periaqueductual
gray matter

Midbrain

Internal
capsule

Pons

Medulla

Spinal cord

Figure 1.1 Two major ascending pathways for nociceptive processing: the spinothalamic
pathway is shown on the left, and the spinomesencephalic pathway is shown on the right
(Adapted from The perception of pain. In Kandel, Schwartz & Jessell (Ed.), Principles of Neural
Science).

1.3.1.1 Spinothalamic and Spinomesencephalic Pathways
The spinothalamic pathway primarily carries nociceptive information about intensity,
location and temperature, and is considered to be the major ascending central pathway for pain
perception (Dennis and Melzack, 1977; Willis et al., 1985). Electrical stimulation of this tract at
the level of the spinal cord or thalamus often evokes pain sensation in humans (Mayer et al.,

6

1975). The primary origin of the spinothalamocortical pathway is the dorsal horn of the spinal
gray matter. Two types of neurons reside in laminae I and V-VII and receive information from
peripheral nociceptive neurons. The first type, nociceptive-specific neurons, respond exclusively
to stimuli that are potentially tissue-damaging. The second type, wide-dynamic-range (WDR)
neurons, respond to weak mechanical stimuli such as light touch or movements of a single hair,
and to noxious stimuli. Axons of both these types of neurons in the dorsal horn ascend in the
dorsal column-medial lemniscal tract, where they synape on neurons in the dorsal column nuclei
in the medulla before decussating and ascending in the anterolateral white matter, and
terminating in the ventral posterior lateral and central lateral nuclei of the thalamus. From the
thalamus, projections finally reach the somatosensory cortex. In addition, some nociceptivespecific and WDR neurons of the dorsal horn project to various ventrolateral and medial
thalamic nuclei, as well as multiple levels of the brainstem and midbrain, including reticular
formation nuclei, the superior colliculus, the periaqueductal gray, parabrachial nuclei and
hypothalamus. These projections reflect the diverse functional consequences of the nociceptive
input.
The spinomesencephalic pathway, another major pain pathway, is thought to contribute
primarily to the affective component of pain. This tract contains axons of neurons in laminae I
and V of the spinal cord. In fact many of the same neurons that give rise to the spinothalamic
tract are also the origin of the spinomesencephalic tract. Axons of these neurons in the dorsal
horn cross to the contralateral side of the spinal cord, ascend in the dorsal column-medial
lemniscal tract, where they synape on neurons in the dorsal column nuclei in the medulla before
ascending in the anterolateral white matter and terminating in the mesencephalic reticular
formation and periaqueductal gray matter of the midbrain. From there, some neurons continue as

7

the spinoparabrachial tract and project from the reticular formation to the parabrachial nucleus,
and finally to the amygdala and hypothalamus. The destination of this tract suggests that it is
likely to contribute to the affective dimension of pain. In addition, the tract may be related to the
mechanism of autonomic responses and arousal that accompany pain experience.
1.3.1.2 Parallel Systems Associated with Sensory and Affective Dimensions of Pain
Projections from the thalamus to the cortex are commonly segregated into lateral and medial
pain systems (Albe-Fessard et al., 1985). The lateral pain system, overlapping with parts of the
spinothalamic tract, consists of projections from the ventral posterior lateral and central lateral
nuclei of the thalamus to both primary and secondary somatosensory cortices (Vogt and Sikes,
2000). The medial pain system consists of projections from medial (midline and intralaminar)
thalamic nuclei to the anterior cingulate cortex and amygdala; neurons from these latter areas
project to nociception-regulating centers such as the periaqueductal gray (Vogt and Sikes, 2000;
Sewards and Sewards, 2002).
The lateral pain system is considered to be intimately involved in processing the sensorydiscriminative dimension of pain sensation. Evidence for this comes from single-unit
electrophysiological studies of the physiological response characteristics of neurons in monkey
primary somatosensory cortex (SI). Kenshalo and Isensee (1983) discovered different types of
nocicepetive neurons in SI with similar receptive fields and physiological response
characteristics to neurons in the spinothalamic pathway. The SI neurons, which receive
excitatory input from thalamic nuclei including the ventral posterior lateral nucleus (VPLc),
respond to graded nociceptive temperatures and provide precise information about nocicepetive
intensity (Price et al., 1992). The secondary somatosensory cortex (SII) also contains nociceptive
neurons and can encode the intensity of nociceptive heat stimulation (Dong et al., 1989). Human

8

studies have shed further light on the functional role of lateral pain system. For example, in one
PET study, Hofbauer et al. (2001) demonstrated that hypnotic modulation of the perceived
intensity of a noxious heat stimulus significantly changed pain-evoked activity within SI,
suggesting that the sensory dimension of pain is encoded in SI.
In contrast to the lateral pain system, the medial pain system includes a number of
interconnected areas within the limbic system. It is these limbic regions which are thare thought
to process the affective dimension of the pain experience. It is known, for example, that ablations
of anterior cingualte cortex (ACC) abolish affective responses of noxious stimuli, while sensory
localization remains intact (Vogt and Sikes, 2000). Nociceptive neurons in ACC have large
receptive fields that usually include the whole body and exhibit the capability to monitor the
overall state of the body. Human neuroimaging studies have found that pain unpleasantness is
mainly encoded in the posterior portion of ACC. For example, hypnotic modulation of the
unpleasantness of experimental heat pain alters activation in ACC but not in SI (Rainville et al.,
1997).
The lateral and medial pain systems, however, do not include all the brain regions that appear
to be involved in the multi-dimensional processing of pain. Besides somatosensory and anterior
cingulate cortical areas, there are insular, posterior parietal, medial and dorsal lateral prefrontal
cortices also activated in neuroimaging studies of pain (Peyron et al., 2000; Symonds et al.,
2006).
1.3.2 The Neuromatrix Theory of Pain
The gate control theory highlights the role of the spinal cord in pain transmission and its
modulation. The neural signals in the dorsal horn that carry information about peripheral input
can increase or decrease the flow of impulses to higher levels of pain processing centers.

9

However, pain is a multidimensional experience and to understand its neural basis, it needs to go
beyond the level of the spinal cord and into the brain where supraspinal regions play very
important roles in pain processing. Recent neuroimaging studies on empathy report that
perceiving pain of others leads to personal distress and activates many brain regions known to be
associated with processing nociceptive stimuli, such as anterior insula, anterior cingulate cortex
and periaqueductal gray (Singer et al., 2004; Cheng et al., 2007). Such findings highlight the
importance of supraspinal mechanisms in generating and modulating pain experience. In addition,
the gate control theory cannot fully explain such phenomena as chronic pain, or phantom limb
pain.
The neuromatrix theory of pain proposed by Melzak (1990) suggests that the
multidimensional experience of pain is produced by characteristic “neurosignature” patterns of
nerve impulses generated by a widely distributed and intrinsically interactive neural network in
the brain. The conceptual model that has emerged from this theory includes four main precepts.
First, the widely distributed neural network, the neuromatrix, includes the neural circuits
between the thalamus and cortex as well as between the cortex and limbic system. Second, the
spatial distribution and synaptic connections of the neuromatrix are initially determined
genetically and are modified by sensory inputs. Third, the neural circuits in the neuromatrix
allow parallel processing of cognitive, sensory and emotional components of pain. The outputs
of parallel processing then converge in the neuromatrix. Parallel processing and synthesis of
nerve impulses generates a characteristic pattern, the neurosignature, of the neuromatrix. Finally,
the neurosignature projects to the brain areas, the sentient neural hub, which convert the flow of
neurosignature into the continually-changed flow of awareness (e.g. pain perception), stress-

10

regulation programs (e.g. cortisol level and immune system activity) or the experience of
movements.
The neuromatrix pain theory helps to account for pain as a multidimensional experience
influenced by multiple stressors. The theory proposes that the neurosignature of pain experience
is determined by the synaptic structure of the neuromatrix and is modulated by sensory inputs.
Thus, the pain experience must be generated in the brain and influenced by the interaction of an
array of neural circuits such as emotion-related brain areas and sensory signaling systems.
1.3.3 Brain Regions of Pain Processing
The neuromatrix pain theory predicts that there must be many different brain areas involved
in pain experience. Because our pain experience includes sensory discriminations, arousal,
cognitive evaluation, affect and motor responses, it makes sense that extensive networks of
cortical and subcortical areas participates in pain processing.
With the development of neuroimaging techniques, it has become increasingly clear that widelydistributed and highly-interconnected brain regions in humans are invovled in generating the
subjective pain experience. In a meta-analysis study reviewing functional imaging of brain
responsive to pain, Peyron et al. (2000) described increased brain activity in a number of brain
areas, including primary and secondary somatosensory (SI and SII), anterior cingulate (ACC),
insular, posterior parietal and medial prefrontal cortices and thalamus. Several of the studies in
the meta-analysis also observed increased activation in the cerebellum, striatum, and
periaqueductal grey. Commonly, SII, insula, ACC, and thalamus are sufficient to generate the
perception of pain to noxious stimuli (Peyron et al., 2000; Apkarian et al., 2005).
1.3.3.1 Brain Regions Associated with the Sensory Dimension of Pain

11

Cortical areas SI and SII contain nociceptive-specific neurons and wide-dynamic-range
(WDR) neurons with similar physiological characteristics to the neurons in the thalamus and the
dorsal horn of the spinal cord.

Electrophysiological studies of SI cortical neurons in the

anesthetized macaque monkey have characterized the responses to a series of noxious
mechanical and thermal stimuli. Response characteristics of SI nociceptive neurons reveal that
these neurons have either restricted contralateral receptive fields or very large receptive fields
covering most of the body’s surface (Kenshalo and Isensee, 1983). The restricted contralateral
nociceptive neurons provide information about the localization, intensity and temporal attributes
of a noxious stimulus. This is true in humans as well: PET studies have found that increases in
regional cerebral blood flow (rCBF) in SI to painful heat stimuli are correlated with the intensity
(temperature) of stimuli (Coghill et al., 1993). A broad region comprising the secondary
somatosensory cortex (SII), neighboring area infraparietal 7b and the restroinsular cortex also
contain nociceptive neurons that encode the intensity of nociceptive heat stimuli (Dong et al.,
1989). The most consistant location of pain-related activity across PET and fMRI studies is
located in this broad cortical region (Peyron et al., 2000). In a number of imaging studies, painrelated activations are reported as a focus of increased cortical blood flow (CBF), which overlaps
on both posterior insular and SII cortices. In addition, the distribution of increased CBF in the
posterior insular/SII cortices is bilateral during painful stimuli (Treede et al., 1999; Symonds et
al., 2006).
1.3.3.2 Brain Regions Associated with the Affective Dimension of Pain
The anterior cingulate (ACC) and the anterior insular cortices are generally considered to
encode the affective dimension of pain processing (Peyron et al., 2000). The ACC is part of the
medial pain system and the nociceptive neurons in the ACC have large receptive fields covering

12

the whole body’s surface. The characteristics of these neurons may reflect the general functions
of the medial pain system in the cortex: to monitor the overall state of the body and to process
the affective components of pain. The role of ACC in the affective dimension of pain is
supported by imaging studies. For example, hypnotic modulation of pain unpleasantness (but not
intensity) altered activation in ACC, but not in SI (Rainville et al., 1997). Tolle et al. (1999)
extended the results of this PET study, using a regression analysis to analyze the relationship of
noxious heat-related regional blood flow (rCBF) increases and experimental pain parameters,
including detection of pain, encoding of pain intensity, and pain unpleasantness. They found that
rCBF in the posterior portion of ACC was positively correlated with pain unpleasantness. A
recent fMRI study showed that even without direct noxious stimulation, imaginary pain
sensation can lead to ACC activation increases (Ogino et al., 2007). In this study, subjects were
instructed to imagine pain in their own body while they were shown pictures of painful events, or
were presented with pictures inducing emotions of fear. Comparing cerebral hemodynamic
responses during the imagination of pain with those to emotions of fear, the researchers found
that most pain-relevant brain areas were activated, including the ACC, right anterior insula,
cerebellum, posterior parietal cortex, and secondary somatosensory cortex region. The ACC is
activated during both imaginary pain and when viewing images evoking emotions of fear. This
suggests that the ACC is likely engaged in both pain and emotional experiences.
The insular cortex receives direct projections from the medial thalamic nuclei and from the
ventral and posterior medial thalamic nucleus. In addition, the anterior portion of the insular
cortex is reciprocally connected with the amygdala (a region associated with emotions). Neurons
in the insular cortex process information on the internal state of the body (Craig, 2003).
Specificcaly, the anterior insular cortex responds to introspective awareness and subjective

13

feeling states (Critchley et al., 2004). Patients with lesions of anterior insular cortex do not
display corresponding emotional responses to the pain, when perceiving noxious painful stimuli
(Greenspan et al., 1999). This suggests that the involvement of the anterior insula in pain
processing is related to the affective dimension of pain.
1.3.3.3 Supraspinal Modulation of Pain Transmission
Investigation of pain-modulation pathways began with the 1969 observation that direct
stimulation of the periaquductal grey (PAG) in the rat could powerfully suppress pain perception
(Reynolds, 1969). The brain regions to which the PAG connects, including the anterior cingualte
cortex (ACC), insula, amygdala, and rostral ventromedial medulla (RVM), form the descending
pain modulation pathways. The RVM receives information from the PAG and serves as an
integral relay in descending modulation of nociception, projecting to the origin of ascending pain
pathways in the dorsal horn of the spinal cord. The RVM has two classes of pain-modulating
neurons: on-cells and off-cells. On-cells are directly inhibited by endogenous opioids and they
facilitate pain transmission. In contrast, off-cells are inhibited by on-cells and they inhibit pain
transmission (Porreca et al., 2002). The existence of on- and off-cells within the RVM raises the
possibility that pain can be modulated by various psychological conditions. For example,
descending pathways from cortical areas could either facilitate or inhibit nociception transmitted
from the spinal cord. Recent fMRI studies do, in fact, indicate that cingulate and frontal cortices
have an effect on the PAG and possibly modulate pain during distraction (Valet et al., 2004). A
particularly compelling fMRI study in which subjects were given in real-time biofeedback data
based on fMRI signal changes showed that both healthy individuals and chronic pain patients can
be trained to control their own activity in the rostral ACC and to modulate their pain experience

14

(deCharms et al., 2005). These studies clearly recognize the role of the cortical regions and PAG
in modulating pain perception.
1.4 The Neural Mechanisms of Emotions
Charles Darwin’s view that many of the different emotions were similar across species laid
the groundwork for the modern field of affective neuroscience. He compared countless sketches
of animals and people in different emotional states and proposed that emotional expressions in
humans are “patterns of action” and that there are a limited set of basic emotions across species
and cultures. These ideas prompted the use of animals in research on human emotions. Later, in
1884 the James-Lange theory proposed that emotions are the experiences of sets of body changes
that occur in response to emotive stimuli, and that different patterns of body changes determine
different emotions (see reviews in Dalgleish, 2004). The most obvious sign of emotional arousal
in animals, including humans, is activity in the autonomic system; e.g. increased or decreased
heart rate, cutaneous transduction and gastrointestinal motility. The James-Lange theory is still
influential in contemporary affective neuroscience, especially in arguments that bodily feedback
modulates emotional experience and that emotional responses can be at least partly distinguished
on the basis of autonomic activity.
1.4.1 Neuroanatomical Theories of Emotions
The first substantive theory of the brain mechanisms of emotions was proposed by Cannon
and Bard in 1931. They investigated the effect of brain lesions on the emotional behavior of cats
and found that removal of the neocortex caused cats to make sudden and ill-directed anger
attacks (Bard, 1937). They argued that the hypothalamus is involved in the emotional responses
and that removal of the neocortex breaks the hypothalamic circuit from top-down control. Their
work suggested that the hypothalamus is a critical center for coordinating emotional behaviors.

15

Around the same time, James Papez (1937) proposed that the central neural circuits, now known
as the ‘Papez circuit’, were responsible for emotional behaviors. He included both the
hypothalamus and cingulate cortex in his proposed emotional circuits, and demonstrated that the
hypothalamus and cingulate cortex were interconnected via projections from the anterior
thalamus and the hippocampus. Papez proposed that emotional information arrived at the
thalamus and diverged into two streams, one of them directed to the cortex (upstream), and the
other directed to the hypothalamus (downstream). The upstream pathway transmitted
information to the sensory cortices, especially to the cingulate region. Through this pathway,
emotional information was converted into perceptions, thoughts and memories. The pathway
extended from the cingulate cortex to the hippocampus and, via the fornix, to the mammillary
bodies of the hypothalamus. The downstream pathway projected from the thalamus directly to
the mammillary bodies of the hypothalamus, completing the circuit and integrating the work of
Bard (1937). The ‘Papez circuit’ accounted both for generation of emotions and the top-down
control of emotional responses.
Over time, the neural circuits for the regulation of emotional responses have come to include
more brain regions. Kluver and Bucy (1939) showed that bilateral lesions of the medial temporal
lobe in rhesus monkeys led to a set of abnormal emotional behaviors (Kluver-Bucy syndrome),
including bizarre oral behaviors, hyperactivity and hypersexuality. It was later shown by
Weiskrantz (1956) that bilateral lesions of the amygdala were sufficient to induce the emotional
disturbance of Kluver-Bucy syndrome. A decade later, MacLean proposed the limbic system
model based on Papez’s and Bard’s ideas and integrated them with the findings from the work of
Kluver and Bucy (MacLean, 1949). This model included many of the components of the ‘Papez
circuit’, such as the thalamus, hypothalamus, hippocampus and cingulate cortex, as well as other

16

important structures such as orbital and medial prefrontal cortex and amygdala. McLean
separated the emotional brain into two parts: (1) the ancient reptilian brain (striatal complex and
basal ganglia), which he proposed is the basis of primitive emotions such as fear and aggression;
and (2) the old mammalian brain (limbic system), which in his view enhances primitive
emotional responses and elaborates social emotions. Finally, in McLean’s model, the new
mammalian brain (neocortex) allows for emotions to interface with cognition and in addition
exerts top-down regulation of emotional responses.
1.4.2 Brain Regions Associated with Emotions
In the past ten years, assumptions regarding the neuroanatomical structures involved in
emotions have been tested using neuroimaging techniques in humans. PET and fMRI studies
have begun to describe the key emotion-related brain regions, such as the anterior cingulate,
insular and medial prefrontal cortices, and the amygdala (Phan et al., 2002; 2004).
The anterior cingulate cortex (ACC) is important for integrating attention and emotional
information, and for regulating emotional behaviors. In particular, the ACC may mediate arousal
responses accompanying emotional behaviors. There is evidence that rostral ACC projects to
autonomic areas such as hypothalamus and brain stem, and with orbitofrontal, insular and medial
temporal regions that also project to these homeostatic centers (Barbas et al., 2003). Direct
stimulation of rostral ACC in animals and humans evokes autonomic responses. PET and fMRI
studies have reported abnormalities in ACC activation in psychiatric patients with obsessivecompulsive disorders, post-traumatic stress, and depression, compared to healthy individuals
(Devinsky et al., 1995; Rauch et al., 1994; Semple et al., 2000; Drevets, 2001). The subgenual
cingulate cortex (sACC) appears to be associated with sadness. Approximately 46% of studies
that induced sadness in the experimental protocol reported that sACC is activated (Phan et al.,

17

2002), and it appears that this region is active whether sadness is induced using either recallgenerated sadness or autobiographical scripts (Reiman et al., 1997; Lane et al., 1997; Liotti et al.,
2000). In PET imaging studies, Mayberg et al. (1999) addressed the potential mechanism
mediating top-down regulation of mood disorders and found that sadness was not only
accompanied with increased activation in sACC and the anterior insula, but also with decreased
activation in the right dorsolateral prefrontal and inferior parietal cortices. These same
investigators also discovered that recovery from depression involved the same regions in a
reversed activation pattern, with decreased limbic metabolism and increased cortical metabolism.
The insular cortex projects to brain regions such as the amygdala and hypothalamus that play
critical roles in regulating autonomic activity (Augustine, 1996). Activation in the insular cortex
has been reported in many neuroimaging studies in which emotions have been manipulated,
including when emotional recall is used and when emotions are induced by viewing pictures
(Phan et al., 2002; Lane et al., 1997). There is also empirical evidence that right anterior insula
activity reflects introspective representation of bodily responses which is crucial to subjective
emotional states arising from autonomic visceral responses (Critchely et al., 2004).
The amygdala also appears to be critical for processing emotions. Neuroimaging studies have
shown that the amygdala responds to fearful faces (Young et al., 1995; Morris et al., 1996;
Philips et al., 1997) and is important for recognizing fearful facial expressions. In addition,
recordings of neurons in the amygdala of the monkey reveal that a population of neurons located
in the basal accessory nucleus of the amygdala responds to human or monkey faces (Leonard et
al., 1985). It has been suggested that the damage of these neurons would cause the deficits in
emotional response (Leonard et al., 1985). Activity in the amygdala is susceptible to the topdown control. For example, it has been shown that attention can modulate amygdala activity

18

(Pessoa et al., 2005). Amygdala activation has been found in response to stimuli other than
fearful faces, including aversive pictures (Taylor et al., 1998), sad faces (Blair et al., 1999), and
even recall of pleasant pictures (Hamann et al., 1999). Thus, it is possible that the amygdala may
respond to general emotional stimulus salience, regardless of emotional valence.
The relationship between the frontal lobe and emotions is perhaps best illustrated by the
classic case of Phineas Gage. In 1848, Gage, a railroad worker, survived an explosion which
caused a large iron rod to be completely propelled through his head, destroying large regions of
his frontal lobes. The injury dramatically affected his personality and behavior. It is now known
that the prefrontal cortex, especially medial prefrontal cortex (mPFC), has an important role in
emotional processing (Phan et al., 2002). The activation of mPFC is engaged in processing
emotions induced by emotional films, pictures and recall (Lane et al., 1997; Reiman et al., 1997).
Positive and negative emotions, such as happiness, sadness and disgust, and the mixture of
emotions all can activate the mPFC.
1.5 The Interaction of Pain and Emotions
1.5.1 Emotions Influence the Perception of Pain
Numerous behavioral experiments have reported that pain experiences are altered by
emotional states. In general, negative emotions increase the perception of pain intensity and
decrease tolerance to pain, while positive emotions attenuate pain intensity, raise pain thresholds,
and increase tolerance. For example, a positive mood induced by humorous film clips increased
cold-pressor pain tolerance, while a negative mood induced by film clips of the holocaust
decreased it (Weisenberg et al., 1998). Meagher et al. (2001) found that male subjects who
viewed slides that provoked feelings of fear or disgust were more sensitive to cold pain, but less

19

sensitive to pain when they watched erotic slides. It has also been reported that perceived pain
intensity is enhanced by pain-relevant anxiety (Ploghaus et al., 2001).
1.5.2 Neural Substrate Correlated with the Interaction between Pain and Emotions
Both emotion and pain are multidimensional contrusts that contain motor, valence, sensory
and physiological components. They are considered to be parts of a large motivational system
that promotes survival (Mollet and Harrison, 2006). In the motivational system, there are
appetitive and defensive sub-systems guiding individuals toward survival-needed stimuli (e.g.
food and water) and warning individuals away from harmful stimuli (e.g. injuries). Emotions are
often described in terms of their place on a range of two main components: valence (from
negative to positive) and arousal (from calm to excited). Positive emotions can be thought of as
the result of appetitive system activity, and negative emotions as a result of defensive system
activity. The level of arousal can be thought of as the degree of activity of the motivational
system. Because pain comprises emotional, sensory and cognitive components, emotion is
inherently part of pain. Pain sensation can be considered to be generated in the defensive system
which motivates organisms to avoid noxious stimuli. The integration of pain and emotion may
account for a wide range of behavioral and physiological changes. It would therefore not be
surprising if a wide range of brain regions were involved in the interaction between pain and
emotion. Further, there should be either some overlap in the neurocircuits of these two states, or
a close association between them. Several limbic structures are potentially associated with the
interaction between pain and emotion.
The cingulate cortex: The medial pain system includes projections from the midline and
intralaminar thalamic nuclei to the anterior cingulate cortex (Vogt and Sikes, 1999). Recently,
Vogt reviewed the linkages between pain and emotion in the cingulate cortex by analyzing the

20

results of 40 human imaging studies in which noxious stimuli were presented, and results of 23
studies in which simple emotions were experienced (Vogt, 2005). His findings include the
following: (1) both noxious thermal stimulation of the skin and fear induced by events, objects or
memories with negative valence activate the anterior mid-cingulate cortex (mACC/BA24); (2)
both happiness and pain induced by balloon distention of the distal esophagus activate a region
within the posterior anterior cingulate cortex (pACC); (3) hypnotic amplification of
unpleasantness during noxious stimulation enhances activity in the caudal part of pACC. These
findings suggest that during some painful states in which both pain and emotion are experienced,
a portion of the cingulate cortex including the pACC and the mACC may be involved in their
integration. Mu-opioid receptors are found in a number of regions in the brain, including both the
ACC and the amygdala, which are shown to release endogenous opioids that bind to µ-opioid
receptors under sustained pain stimuli (Zubieta et al., 2001).
The Insular cortex: Insular cortex plays a role in diverse functions such as emotions,
homeostasis, and self-awareness (Craig, 2002; Critchley et al., 2004). Both pain and emotion
represent important changes in the internal states of an organism. Therefore, it is possible that the
link between emotion and pain may be associated with the activity in the insular cortex. Damasio
et al. (2000) reported in a PET study that processing emotion activated regions of the insular
cortex. In their experiments, 41 normal subjects recalled and re-experienced personal life
episodes marked by sadness, happiness, anger or fear. The results indicated bilateral, but
asymmetric, activations in anterior/middle insular cortex during either sadness or anger, as well
as positive activations in right anterior/middle insular cortex during either happiness or fear. In
another study, the activity in the bilateral anterior insular cortex of a neuropathic pain patient was
increased by unpleasant odors (Villemure et al., 2006). This identified activation pattern in

21

anterior insula was close to the activation pattern of sadness found in the experiment of Damasio
et al. (2000). Evidence from these studies suggests that the insular cortex may be an important
region involved in both emotional and painful experience.
The amygdala: The amygdala may also integrate neural activity associated with negative
emotions and pain. In a PET study, Zubieta and colleagues (2001) examined the function of the
opioid system in healthy human subjects undergoing sustained muscle pain. Significant
activation of the µ-opioid receptor system was detected in selected volumes of interest in the
amygdala ipsilateral to the painful stimulus, indicating a regional release of endogenous opioids
during sustained pain. In another PET study, Zubieta et al. (2003) reported that sustained sadness
was associated with significant deactivation of µ-opioid neurotransmission in the amygdala.
These lines of evidence suggest that in addition to the amygdala’s well-documented role in
affective processing, it is involved in the µ-opioid-mediated sensory responses.
The frontal lobe: The medial prefrontal cortex (mPFC) also may be associated with the
interaction between emotion and pain. An fMRI study in rheumatoid arthritis patients found that
applying pressure to arthritic joints provoked joint pain and led to increased activation in the
bilateral MPFC (Schweinhardt et al., 2008). Moreover, the increased activation in the MPFC was
significantly positively correlated with the patients’ Beck Depressive Inventory (BDI) scores.
The right ventrolateral prefrontal cortex (VLPFC) may be involved in emotional regulation of
pain by driving top-down pain inhibitory pathways. This region, for example, was identified in
an fMRI study in which religious beliefs decreased pain perception. Among practicing Catholics,
an image of the Virgin Mary induced context-dependent analgesia during acute electrical pain
stimuli, concurrent with increased activation in the right VLPFC (Wiech et al., 2008).
1.6 Background to the Methodology

22

Until twenty year ago, pain was investigated primarily at the peripheral and spinal cord levels.
Studies mainly relied on animal models, in which ablation methods, pharmacological
manipulation and electrophysiological methods were employed. Functional neuroimaging has
fundamentally changed our knowledge about pain sensation and led to the identification of brain
regions linked to pain sensation in humans. Because pain is a subjective experience and can
change under different circumstances, non-invasive techniques such as functional magnetic
resonance imaging (fMRI) allow us to investigate the brain activity of conscious subjects and to
record functional activity across the entire brain simultaneously. The experiments in the
dissertation employ fMRI methods to investigate functional networks for altered pain experience
by sad mood. First, task-driven fMRI methods are used to identify brain activation patterns
when pain is affected by sadness. Second, resting-state fMRI methods are used to investigate the
intrinsic connectivity within brain networks for pain and emotion. Third, path analysis is applied
to the fMRI datasets to investigate the functional connectivity of brain networks that underlie
altered pain perception during sadness.
1.6.1 Task-driven fMRI and Data Analysis
Most fMRI experiments examine brain activity by measuring the blood-oxygenation-level
dependent (BOLD) signal, which relies on the different magnetic properties of oxygenated and
deoxygenated hemoglobin (Ogawa et al., 1990). The BOLD signal is correlated with neuronal
activity, as local neuronal activity increases metabolic requirements. Logothetis et al. (2001)
presented empirical evidence that local field potentials in the visual cortex of monkeys were
significantly correlated with the BOLD hemodynamic response. In that landmark study,
intracortical electrophysiological recordings were made while simultaneously recording the
BOLD signals with fMRI.

23

Task-driven fMRI experiments detect regional changes in the BOLD signal as neurons in
specific brain areas respond to the designed tasks. For example, in this dissertation study, a taskdriven fMRI study was designed to deliver pain, either alone, or accompanied with visual stimuli
to investigate how sadness influences pain perception. Good experimental design is necessary in
functional MRI studies and several issues need to be considered including: (1) the specific
dependent variables to be measured (e.g. brain activation under different tasks); (2) whether a
between-subject or within-subject design is more appropriate; (3) the implementation of
adequate control conditions; (4) control of confounding factors (e.g. induced emotions other than
sadness); and (5) the signal-to-noise ratio necessary to statistically detect the BOLD signal. Taskdriven fMRI studies do not “draw pictures” of absolute BOLD signals, but analyze the percent
change in the signal by comparing task periods to an arbitrary baseline of non-task periods. Brain
regions that are active and that potentially participate in performing the task are identified on the
basis of a statistically-significant difference in fMRI signal intensity during the task compared to
the baseline. In this dissertation study, there are three different tasks: (1) painful electrical shocks;
(2) mood induction with emotionally-evocative pictures; (3) painful electrical shock
accompanied by emotional pictures. Non-emotional pictures are used during the baseline periods
to control for activation associated with merely watching and paying attention to colorful
pictures. To evaluate brain activation in a task condition compared to a baseline condition, most
fMRI experiments use either an event-related design or a block design, or a mixture of the two.
Block designs are most useful for discovering the magnitude of neural activity in one condition
versus another, especially if the response to the task is slow, as in responses to electrical shocks
and presentation of affective pictures. Block designs are powerful for detecting significant fMRI

24

activity, and minimizing low-frequency noise, and the design and analysis are relatively
straightforward.
The procedure of a task-drive fMRI experiment is shown in the following chart (Figure 1.2).
Functional brain imaging data is collected on individual participants. The experimental design is
based on the number of experimental trials, the number of blocks in each run, the length of each
block and the number/duration of stimuli during each block. Once the experimental design is
finalized, participants are recruited to collect functional and anatomical datasets in MRI. The
functional datasets of each individual are preprocessed to convert 2D images into 3D datasets,
align separate slices to the same temporal origin, correct head movements, and eliminate outliers.
Assuming the hemodynamic response function (HRF) has a fixed shape in any active brain
region regardless of stimulus type, we can focus on the differences of the magnitude of the HRF
in active brain regions. Regression analysis is applied to an individual subject’s dataset to
identify the significant brain activation during different tasks. Group analysis such as ANOVA
can then be used to identify significantly activated brain regions across subjects. Finally, regions
of interest analysis or connectivity analysis can be used in post-processing.

25

Experimental Design

MRI Scanning

Data Pre-Processing

Individual Subject Data Analysis

Group Data Analysis

Data Post-Processing
Figure 1.2 Flow charts of a task-driven fMRI experimental procedure.
1.6.2 Resting-State fMRI and Data Analysis
Traditional task-driven fMRI studies detect increased brain activity in response to specific
experimental tasks. The signal that is analyzed in task-driven fMRI studies only accounts for a
small amount of body metabolism, approximately 5% or less. The resting brain consumes 20%
of the body’s energy. This raises the question of why the brain consumes so much energy in the
resting state. One explanation comes from the observation that most of the energy is used to
support ongoing neuronal signaling (Fox et al., 2005). Thus, the spontaneous BOLD signals,
which indirectly measure neuronal activity based on the brain metabolism, in the resting human
brain could be meaningful. It has been found that spontaneous BOLD activity in the left
26

somatomotor cortex is specifically correlated with spontaneous fluctuations in the right
somatomotor cortex and with medial motor areas in the absence of motor behavior (Fox et al.,
2005). This empirical observation suggests that the spontaneous BOLD activity is not random
noise, but reflects the functionality of the resting brain. Detecting brain activation patterns during
resting states could provide insights into the intrinsic functions of the brain.
Based on the assumption that the components of a neural network are functionally connected,
the BOLD signal fluctuations of these components can be reasonably assumed to be correlated in
the resting state. This assumption is the basis of resting-state fMRI. Functional imaging studies
of the resting brain have been conducted for many brain systems, including somatomotor, visual,
auditory, episode memory and attention systems (see reviews in Fox and Raichle, 2007). These
studies have shown that brain regions with similar functionality tend to have correlated
spontaneous BOLD activity. Many resting-state fMRI studies have found that a particular set of
brains regions is typically more excited at rest than during cognitive task performance (Raichle et
al., 2001; Greicius et al., 2003). This pattern of activity is called the “default mode” brain
network, and includes both positive and negative correlations of activity among separate brain
regions. The task-negative regions are distributed in the medial prefrontal, lateral parietal, and
posterior cingulate/precuneus cortices. Task-positive regions are distributed in the intraparietal
sulcus, the frontal eye field region and the middle temporal region.
Resting state fMRI methods have been used to investigate the neural mechanisms that
underlie pain. Baliki and colleagues (2008) investigated how chronic back pain disrupts the
default mode network during the resting state. Using the brain regions of the default mode
network as the seed regions in a correlation analysis, they found that correlation maps of healthy
controls were very close to the default mode network described by Raichle et al. (2001) and

27

Greicius et al. (2003). However, chronic pain patients showed much smaller cortical areas with
negative correlations and larger areas with positive correlations than healthy controls. The
reduction in the total brain area exhibiting anticorrelation suggests a disruption of the balance of
the positive and negative correlations in the default mode network and impairment of cortical
functional connectivity in chronic pain patients. The disruption of the default mode network in
the chronic pain patients may underlie the cognitive and behavioral impairments that accompany
chronic pain. In another study, Cauda and colleagues (2009) explored the thalamocortical
connectivity in a group of patients suffering from peripheral neuropathic pain by using a resting
state fMRI paradigm and predefined regions of interest (i.e., primary somatosensory cortex,
ventral posterior lateral thalamic nucleus, and medial dorsal thalamic nucleus). Compared with a
group of healthy subjects, the patients showed decreased temporal correlations between
predefined regions of interest, indicating a decreased resting state functional connectivity
between the thalamus and the cortex. These data support the authors’ interpretation that chronic
pain can disrupt thalamic feedback during the resting state.
Both the scanning protocol and pre-processing methods used in resting fMRI studies are
similar to those in task-driven fMRI studies. Common techniques used to identify spatial patterns
of spontaneous BOLD activity include simple correlation analysis and independent component
analysis (ICA). Simple correlation analysis is to extract the BOLD signal from a defined seed
region and determine the temporal correlation between the extracted signal and signals from
other brain regions. This technique is straightforward and the results are easy to interpret. In
contrast, ICA does not require defining a seed region. The algorithm of ICA decomposes the
entire BOLD dataset into independent components which are maximally independent statistically.
Each component represents either a meaningful spatial pattern of spontaneous brain activation or

28

noise. Noise components are automatically separated, but researchers must determine which
components reflect noise and which reflect functional anatomical networks.
1.6.3 Functional Connectivity of the Pain Network
1.6.3.1 Path Analysis
Interregional Pearson product-moment correlation coefficients historically have been used to
determine which brain regions are functionally correlated during a specific experimental task.
However, interpretations based on pair-wise correlations are complicated when multiple
neuroanatomical systems are involved. Path analysis was therefore proposed as a method to
quantify the interaction among many brain regions (McIntosh et al., 1992). Path analysis uses
information about the anatomical connections and correlation coefficients among brain regions
to determine the functional connectivity in a given experiment. It attempts to find causal
relationships among brain regions. The application of path analysis to fMRI datasets is based on
the assumption that correlation patterns between two brain regions are due to either common
influence on both regions or to direct anatomical connections between them. If two regions are
functionally linked in a specific task, the correlation coefficient of their activity will be large.
The correlation between two brain regions is expressed as the sum of the compound paths
connecting the two regions. Mathematically, the goal of path analysis is to obtain a set of path
equations to minimize the difference between the observed covariances and those implied in the
solution of equations. The analysis uses a series of successive “guesses” to find the best-fitting
model of the observed fMRI dataset. Increased path coefficients from region A to region B under
different experimental conditions will indicate increased strength of functional connectivity
between those regions. In other words, activation in region A would predict more increased
activation in region B while holding all other relevant regional connections constant.

29

1.6.3.2 Anatomical Connections of Brain Networks for Pain and Emotion
To apply functional connectivity techniques, it is necessary to construct hypothetical models
of functional connectivity based on what is known about the anatomical connections among
target brain regions. Fortunately, the ascending spinothalamocortical pathway has been
extensively investigated in both animal lesion studies, and electrophysiological and neural
imaging studies for humans (see reviews in Price, 1999). In this dissertation, we proposed a
hypothetical model for pain processing based on known connections in the pain network (Figure
1.3). Projections originating from the dorsal horn of the spinal cord reach the central targets in
brain stem sites, including the periaqueductal gray (PAG), continue to the ventrocaudal part of
the medial dorsal nucleus (MDvc) and the ventroposterior lateral (VPL) nucleus of thalamus, and
finally terminate in the cortical somatosensory areas (i.e., SI and SII). SI and SII are reciprocally
connected, and projections from these two areas continue to the cortical limbic structures
including insular and cingulate cortices (Manzoni et al., 1986; Conti et al., 1988). Anatomical
evidence from monkey studies suggests that reciprocal connections exist between SII and the
parietal lobe/ BA7b (PPC), and that there are projections from the parietal lobe to the cingulate
gyrus/BA 23 (PCC) as well as the insular cortex (Neal et al., 1987). The insula also receives
direct afferents from SII (Mesulam and Mufson, 1982). Within the cingulate cortex, the posterior
subdivision/BA23 (PCC) has reciprocal connections with the anterior subdivision/BA 24 (ACC),
and both subdivisions have widespread connections within the cortex including the insular region
(Pandya et al., 1981). In addition, the anterior cingulate and the anterior insular cortices receive
direct projections from the ventromedial part of the posterior nuclear complex (VMpo) and
ventrocaudal part of the medial dorsal nucleus (MDvs) of the thalamus (Zhang and Craig, 1997).

30

M1

S1

SMA
S2
ACC

PCC

PPC

INSULA
MDvc
PF

VMpo

VPL

HT
AMYG

PAG
PB

Figure 1.3 Cerebral cortical pathways involved in the pain processing (Price, 2000).
To construct a functional connectivity model of the interaction between pain and emotion,
connections of the pain network mentioned in the previous paragraph are included. In addition,
several other areas and their connections that are part of the brain’s emotional circuitry are
included in the model. Subgenual cingulate cortex (sACC/BA25) is one of the important nodes
for processing sadness; this area has reciprocal connections with other subdivisions of the
cingulate cortex, including anterior cingulate cortex (ACC/BA24) and posterior cingulate cortex
(PCC/BA23) (Pandaya et al., 1981; Cavada and Goldman-Rakic, 1989). Therefore, reciprocal

31

connections between ACC/BA 24 and PCC/BA23, as well as reciprocal connections between
ACC/BA24 and sACC/BA25 are included in the hypothetical model to analyze functional
connectivity of pain and emotion networks. The amygdala/hippocampal complex is another
important node involved in emotional experience and has widespread reciprocal projections with
the cingulate cortex and the periaqueductal gray (PAG). The direct projection from the amygdala
to the PAG has an important role in opioid analgesia (Finnegan et al., 2004). It is therefore
necessary to include the reciprocal connections between amygdala and PAG, as well as
connections between amygdala and cingulate cortex in the same model. The insular cortex is
another important brain region involved in processing emotions and homeostasis. The insular
cortex has reciprocal connections with the amygdala (Mufson et al., 1981). It also has
connections with wide-spread brain regions such as the secondary somatosensory, prefrontal,
temporal, and parietal cortices as well as the PAG (Augustine, 1996). Finally, the medial
prefrontal cortex projects to the anterior cingulate cortex and from there to posterior cortical
regions, as well as to subcortical regions and the amygdala (Price, 2000; Miller and Cohen,
2001).

This projection originating in the prefrontal cortex may be involved in top-down

modulation of pain sensation.
1.7 Research Questions of This Dissertation
Pain research has dramatically advanced in the past 40 years. In the 1970s, we knew almost
nothing about the cortical representation of pain; currently, we know that widely-distributed
cortical regions are involved in pain processing. The integration of pain and emotions has been
addressed in several brain regions including the anterior cingulate, the amygdala and the insular
cortex. However, to our knowledge, few studies explore the effect of negative emotions on pain
perception in a functional brain network. This dissertation focuses on the interaction between

32

pain perception and sadness at the cortical and subcortical level in humans. The first research
aim explores how sadness affects subjective pain in the brain. The second research aim identifies
the functional network in which pain-relevant and emotion-relevant brain regions interconnect
when pain experience is altered by sad mood. Knowledge of these funcitonal networks will help
to understand the mechanisms underlying subjective pain experiences and identify neural circuits
that may be important to understanding the comorbidity of pain and emotional disorders.

33

Chapter 2
An fMRI Study of Pain Perception Modulated by Sad Mood
2.1 Introduction
Pain is both an objective sensory experience that mirrors the intensity of noxious stimuli, and
a subjective one that can range from mildly unpleasant to excruciatingly noxious. Emotional
state can profoundly influence pain perception. Laboratory studies have shown that emotions
induced by self-reflective statements, music, pictures or film-clips with emotional content can
alter pain experience (Willoughby et al., 2002; Tang et al., 2008; Meagher et al., 2001;
Weisenberg et al., 1998). For example, inducing a depressed mood using self-reflective
statements can significantly shorten the time an individual can tolerate keeping a hand
submerged in ice-cold water (Willoughby et al., 2002). Among chronic back pain patients,
experimentally induced depressed mood increases self-reported pain and decreases tolerance of
pain elicited by holding a heavy bag (Tang et al., 2008).
Neuroimaging techniques such as positron emission tomography (PET) and functional
magnetic resonance imaging (fMRI) are increasingly being used to study brain activation in
humans using experimental protocols in which painful stimuli are delivered to participants who
can simultaneously report their subjective pain experience. In such experiments, neural
activation during painful stimulation is distributed widely throughout cortical and subcortical
regions of the brain, including somatosensory, prefrontal, insular and anterior cingulate cortices,
areas that are included in the so-called “pain matrix” (Peyron et al., 2000; Melzack, 1990). There
is now evidence that some of these brain regions spatially overlap those associated with emotion
(Vogt, 2005). For example, regions in cingulate and insular cortices that process pain sensation
are also likely to be involved in both mood disorders and ordinary emotion (Apkarian, 2005;

34

Ressler and Mayberg, 2007; Damasio et al., 2000).

Furthermore, neuroimaging data from

patients with either depressive or chronic pain conditions suggest that the neural substrates of
pain and depression are at least partially overlapping. For example, when painful thermal stimuli
are administered to the right arm, prefrontal and insular cortices show hyper-activation in
patients with major depression, compared with healthy subjects (Bär et al., 2007). When pressure
is applied to a painful joint of rheumatoid arthritis patients, the neural activity in an area of the
medial prefrontal cortex that is often active during the experience of pain is significantly
correlated with their scores on the Beck Depression Inventory (Schweinhardt et al., 2008).
Studies such as these suggest that brain regions relevant to pain and to depression may be
functionally associated with one another. At this point we do not know how neural activity
related to an emotional state could affect activity in brain regions responsive to pain, or whether
such a relationship helps to explain why chronic pain and depression are so often co-morbid.
In this study, we used emotionally evocative pictures to induce a sad mood in healthy female
subjects and compared neural activity in response to painful electric shocks in both emotional
and non-emotional conditions. Our hypotheses were that: (1) subjective perception of pain
intensity would increase during sadness; and, (2) increased pain perception would be
accompanied by significantly altered neural activation in some regions of the pain matrix,
particularly the somatosensory and insular cortices.
2.2 Methods
2.2.1 Subjects
Sixteen right-handed, healthy women (ages 19-22) were recruited from the Department of
Psychology Human Subject Pool at Michigan State University. None of the participants had a
history of chronic pain, substance abuse, hospitalization for psychiatric disorders or brain injury.

35

Handedness was evaluated using the Edinburgh Handedness Inventory (Oldfield, 1971) and an
MRI safety questionnaire was administrated to ensure that participants met inclusion criteria for
imaging. Participants provided written informed consent, and they were free to withdraw from
the experiment at any time. All procedures were approved by Michigan State University’s
institutional review board.
Upon arrival, participants completed two self-report questionnaires to screen for symptoms
of psychiatric disturbance or medical disorders that could affect pain perception: the Beck
Depression Inventory (BDI) (Beck et al., 1996) and the Brief Symptom Index (BSI) (Derogatis,
1992). Cutoff scores for excluding participants were greater than 19 for the BDI and greater than
63 for the BSI. In addition, participants were administered the Affective Intensity Measure (AIM)
(Larsen, 1984) to ensure that they were likely to respond emotionally to the picture stimuli.
Cutoff scores for excluding participants was less than 80 for the AIM.
2.2.2 Experimental Procedure
Painful Stimuli: Acutely painful stimuli were delivered through gold-plated electrodes
applied to the lateral surfaces of the second phalange of the right index finger using a 9-V
®

battery-powered Neurometer

adapted for the MRI environment (Neurotron, Baltimore, MD).

To maximally stimulate small-diameter C fibers, the AC current was set at 5 Hz (Katims et al.,
1997; Symonds et al., 2006).
Pain thresholds and current intensity settings were individualized for each participant using a
standard procedure developed in the laboratory (Symonds et al., 2006). Briefly, the procedure
was as follows. For each participant we first obtained the Current Perception Threshold (CPT)
and Nociceptive Threshold (N-CPT) following the standard procedure described in the manual
for the Neurometer. Pain ratings (5 to 100) of 3-sec pulses delivered at 0.05 mA increments were

36

then collected from each subject. These pain ratings for multiple stimulus intensities were used
to establish a linear stimulus-response function. To equalize subjective pain intensity across
participants, the current intensity corresponding to a numerical rating of 35 was selected for
painful stimulation during imaging. All participants were informed that the painful stimulus
during imaging would be at an intensity somewhere between their ratings of 5 and 60. In reality,
it remained constant throughout all experimental conditions, set for each participant at the
intensity that had been given a rating of 35. Several pulses of painful stimuli of varying intensity
were tested briefly when participants were placed into the scanner to familiarize them with the
experimental environment and to reduce anxiety.
Mood Induction and Ratings: Pictures selected from the International Affective Picture
System (IAPS) (Lang et al., 1993) were presented on an LCD screen in the scanning room.
Thirty pictures with negative valence were specifically used to induce sadness. All pictures had
low-to-moderate arousal ratings (< 6.5). Previous studies in our laboratory have established that
these pictures specifically induce sadness and do not induce other negative emotions such as fear,
disgust or anxiety, which have been reported to influence pain perception (Sudheimer et al.,
2001). Thirty non-emotional neutral pictures (mean valence = 4.85; mean arousal = 3.26) were
used during the control condition. Participants rated both their level of pain perception and their
emotional states on a numerical scale (0 = ‘no pain at all’/ ‘least possible’; 100 = ‘worst pain
imaginable’/ ‘most possible’). Ratings for intensity of both pain and emotions were obtained
before and after each imaging trial to obtain baseline and post-trial measurements.
Imaging Protocol: Participants were placed in the scanner, and a strip of tape was secured
across the forehead to minimize head movement. To familiarize subjects with the fMRI
environment, a 2.5-min practice imaging trial was run. After the practice run, three separate 4.5-

37

min trials were run for each participant (see Figure 2.1). Each trial was divided into alternating
30-sec periods of rest and task. Rest periods for all three trials consisted of presenting a sequence
of five non-emotional (neutral) IAPS pictures (6 sec/each). Stimuli for task periods for each of
the three trials were as follows: Trial 1 (Pain Only) six electrical shocks (3-sec on, 2-sec off);
Trial 2 (Sad Only) five sad pictures presented sequentially, each for 6 seconds; Trial 3 (Sad and
Pain), electrical shocks as in Trial 1 and sad pictures as in Trial 2 administrated simultaneously.

Trial 1: Pain Only
Trial 2: Sad Only

SadPic

SadPic

SadPic

Trial 3: Sad and Pain

SadPic

SadPic

SadPic

Figure 2.1 Block design of fMRI trials: Pain Only, Sad Only, and Sad and Pain. Dotted black
line in Trial 1 and Trial 3 indicates delivery of 6 pulses of electrical shock, each presented for 3
seconds with 2 seconds between pulses.
2.2.3 Image Acquisition
Imaging was performed on a GE 3.0 Tesla scanner equipped with an eight-channel head coil,
*

allowing for whole brain imaging. Each functional scan continually collected T2 -weighted
images of 74 whole brain volumes (matrix size = 64 × 64). The first 4 volumes were excluded to
allow the scanner to reach a steady state. Functional images were acquired using gradient echoo

planar pulse sequences (TE = 25 ms; TR = 3000 ms; flip angle = 90 ; FOV = 240 mm; voxel
size = 3.75 × 3.75 × 4 mm). The whole brain volume was covered by 38 contiguous 4 mm thick
axial slices. An anatomical scan following functional imaging collected 128 T1-weighted axial
o

images (TE = 5ms; TR = 24 ms; flip angle = 90 ; FOV = 240 mm; matrix size = 256 × 256;

38

voxel size = 0.938 × 0.938 ×1.5 mm). These high-resolution Spoiled Gradient-recalled Echo
(SPGR) images were used as an underlay for functional activity maps.
2.2.4 Image Analysis
All images were analyzed using the Analysis of Functional NeuroImages (AFNI) software
package (Cox, 1996). Analysis of fMRI data followed conventional procedures for processing
single subject data, conversion to standard stereotaxic coordinates (Talairach and Tournoux,
1988), group analysis, and region of interest (ROI) analysis. These procedures are briefly
described below.
For each subject, raw functional and anatomical image files (two-dimensional slices) were
converted to three-dimensional AFNI datasets. For functional images, Echo Planar Imaging (EPI)
time series were corrected for inter-slice time shift and in-plane motion artifact. EPI time series
were then processed using spatial smoothing with an 8 mm FWHM Gaussion filter to reduce
random low-frequency noise. EPI time series were normalized as percent signal changes, relative
to the mean BOLD signal in each voxel. A reference function was selected as an ideal predictor
of the hemodynamic response function (HRF) based on the timing of the experimental protocol’s
block design (Figure 2.1). In a multiple regression analyses, the amplitude of the normalized
Blood-Oxygen-Level-Dependent (BOLD) signal was then predicted on a voxel-by-voxel basis
with multiple predictors, including the gamma waveform that best predicted the HRF, a constant
to represent the baseline, the degree of the polynomial in the baseline model, and the motion
correction parameters. General linear tests were also applied to each voxel to generate F-statistic
and t-statistic maps representing changes in BOLD contrast, and pair-wise comparisons for all
experimental conditions were performed. Images for both anatomical and functional time-series
3

were linearly interpolated to volumes with 1-mm voxels, then co-registered and converted to

39

standard stereotaxic coordinates to allow averaging across subjects. The percent signal changes
of each dataset (i.e., participant) were used as input for group analysis, a voxel-by-voxel analysis
of variance (ANOVA).
For whole brain group analysis, ANOVA was used to investigate significant activation in
brain regions during the three trials (Pain Only, Sad Only, and Sad and Pain) with a mixed-effect
two-factor model. Experimental trial (3 levels) was the first factor (fixed effect) and subject (15
levels) was the second factor (random effect). One dataset from the 16 subjects was discarded
from the imaging dataset, because she did not show a difference in pain perception ratings during
the experimental Sad and Pain trial compared with the Pain Only trial. Monte Carlo simulations
were applied to obtain the corrected type I error (i.e., alpha level) and cluster size for the group
analysis of the whole brain. For example, based on the Monte Carlo Simulations at an individual
-5

voxel-detection probability of p ≤ 2 × 10 , the corrected alpha level for the whole brain was
determined to be 0.001, and the clusters considered to be active would have to be at least 125
3

3

mm . Cluster volumes had to be at least 55 mm at a more relaxed alpha value of 0.005. In this
study, alpha value less conservative than 0.005 is not used, so that it is possible to separate
adjacent clusters (e.g. amygdala/hippocampus and occipital/fusiform gyri). A mask was created
using a corrected alpha level threshold and a set of clusters was obtained to use as regions of
interest (ROIs). Average z-scores across fifteen subjects for each experimental condition or
contrast condition were then obtained for each of the ROIs. In addition, the contrast analysis of
(Sad and Pain – Pain Only – Sad Only) was performed to detect significantly increased brain
activation for the condition in which painful and sad stimuli were presented simultaneously
compared to conditions in which they were presented alone. In this contrast analysis, many
potentially meaningful ROIs (e.g., subgenual cingulate cortex, periaqueductal grey, and right

40

amygdala/hippocampal complex) did not survive at an alpha level of 0.001, so a more liberal
alpha level of 0.005 was applied in order to explore these potentially relevant regions.
2.3 Results
2.3.1 Behavioral Data
All sixteen participants completed the experimental protocol and were included in the initial
analysis of behavioral data. Beck Depression Intensity scores and Brief Symptom Index scores,
which were all below the exclusion cutoff, were 3.8 ± 3.4 SD and 23.5 ± 17.5 SD respectively.
Affect Intensity Measure scores were all higher than the cutoff (mean of AIM: 136.4 ± 15.1 SD).
The sad emotion induction protocol successfully induced sadness in the group of sixteen
participants: mean numerical ratings of sadness for the Sad Only trial were 7.0 ± 3.2 before and
47.4 ± 7.1 after (paired t (15) = 5.98, P < 0.0001). These ratings were not significantly different
from those during the Sad and Pain condition (paired t (15) = 1.07, P = 0.2, see Figure 2.2).
In further analyses, data from one participant was excluded because she did not show
increased pain perception in the Sad and Pain trial. For the remaining 15 participants, the mean
rating of perceived pain intensity during the Pain Only trial was 37.9 ± 5.6. As expected, during
both the Pain Only and the Sad and Pain trials, pain intensity was rated significantly higher for
the task periods (pain stimulation) compared to baseline (paired t (14) = 5.8, P < 0.001; paired t
(14) = 6.5, P < 0.001, respectively). Finally, pain was rated as more intense when painful stimuli
were administered while participants were viewing sad pictures and experiencing a sad mood
than when they were viewing neutral pictures, even though the intensity of the painful
stimululation was identical in both conditions (mean pain rating for Sad and Pain was
51.41±6.67 and mean rating for Pain Only was 37.89 ± 5.63; paired t (14) = 2.6, P = 0.02, see
Figure 2.3).

41

Sadness
70.00

*

60.00
Rating

50.00
40.00

Baseline

30.00

Task

20.00
10.00
0.00
Pain Only

Sad Only

Sad and Pain

Figure 2.2 Mean ratings of sadness before and after each experimental task. The ‘*’ indicates
significant difference in the paired t-test (P < 0.0001)

Pain
70.00

*

60.00
Ratings

50.00
40.00

Baseline
Task

30.00
20.00
10.00
0.00
Pain Only

Sad Only

Sad and Pain

Figure 2.3 Mean ratings of pain intensity before and after each experimental trial. The ‘*’
indicates that pain was rated significantly more intense for the Sad and Pain trial than Pain Only
in the paired t-test (p < 0.001).

42

2.3.2 Neuroimaging Data
2.3.2.1 Brain Regions Activated During Experimental Tasks
Location of brain activation in response to acute painful shocks was consistent with previous
findings from both our own lab (Symonds et al., 2006) and other groups (see reviews in Peyron
et al., 2000). During the Pain Only condition, significant brain activation was observed in the
contralateral

(left)

primary

somatosensory

(SI),

bilateral

posterior

insular/secondary

somatosensory (pIn/SII) and contralateral (left) anterior cingulate (BA 32) cortices (Figure 2.4;
see Table 2.1 for coordinates, volumes of clusters, and mean z scores). These three brain regions
are main components of the brain network activated during acute pain conditions, as summarized
in the meta-analysis of Apkarian (2005). During the Sad Only condition, significantly active
regions were bilateral medial prefrontal cortex (BA9), ipsilateral (right) inferior frontal cortex
(BA 45, 44), bilateral amygdala/hippocampal complex, and bilateral occipital/fusiform gyri
(Figure 2.5 and Table 2.1). Finally, during the Sad and Pain condition significant activation was
located in bilateral pIn/SII and contralateral (left) anterior cingulate (BA 32) cortices. These two
regions of activation were also observed during Pain Only. Significant activation during Sad and
Pain was also observed in the ipsilateral (right) medial prefrontal cortex, bilateral
amygdala/hippocampal complex, bilateral occipital/fusiform gyri and ipsilateral (right) anterior
insular cortex. Except for the insular cortex, these regions also contained significant activation
during Sad Only. Subcortically, during Sad and Pain, significant activation was observed in the
contralateral (left) thalamus and contralateral (left) periaqueductal grey (Figure 2.6, Figure 2.7
and Table 2.1).

43

A

Z
B

≥5

0

C

L

R
≤ -5

Figure 2.4 Brain activation during the Pain Only condition. Images in all three rows left to right:
axial, coronal and parasagittal. Left hemisphere shown on left in axial and coronal views. Green
crosshairs centered on region of interest. Row A: left primary somatosensory cortex (SI); Row B:
bilateral posterior insular/secondary somatosensory cortices (pIn/SII); Row C: left anterior
cingulate gyrus (BA 32). Alpha value, corrected for whole brain, is 0.001. (For interpretation of
the references to color in this and all other figures, the reader is referred to the electronic version
of this dissertation.)

44

A

Z
B

≥5

0

C

≤ -5

Figure 2.5 Brain activation during the Sad Only condition. Images in all three rows left to right:
axial, coronal and parasagittal. Left hemisphere shown on left in axial and coronal views. Green
crosshairs centered on region of interest. Row A: bilateral medial prefrontal gyrus (BA 9); Row
B: right inferior frontal gyrus (BA 45/44); Row C: bilateral amygdala/hippocampal complex.
Circles in the axial image indicate bilateral occipital/fusiform gyri. Alpha value, corrected for the
whole brain, is 0.001.

45

A

B
Z

C

≥5

0

D

≤ -5

Figure 2.6 Brain activation during the Sad and Pain condition. All images are in the coronal
plane. Green crosshairs centered on region of interest. Left hemisphere shown on left. A:
bilateral posterior insular/secondary somatosensory cortices (pIn/SII); B: left anterior cingulate
gyrus (BA 32). C: right medial prefrontal cortex; D: bilateral amygdala/hippocampal complex.
Alpha value, corrected for the whole brain, is 0.001.

46

A

Z ≥5
B
0

≤ -5

C

Figure 2.7 Brain activation during the Sad and Pain condition. Images in all three rows left to
right: axial, coronal and parasagittal. Left hemisphere shown on left in axial and coronal views.
Green crosshairs centered on region of interest. Row A: left periaqueductal grey (PAG); Row B:
left thalamus; Row C: right anterior insular cortex. Alpha value, corrected for the whole brain, is
0.001.

47

Condition

Cluster location

Active
volume
(mm3)
Left SI
479
Pain Only Left posterior insula/SII
2516
(P ≤
Right posterior insula/SII 2164
0.001)
Left anterior cingulate 162
(BA 32)
Bilateral
medial 550
prefrontal (BA 9)
Right inferior frontal (BA 1979
Sad Only 44/45)
(P≤
Left
2555
0.001)
amygdala/hippocampus
Right
2033
amygdala/hippocampus
Left occipital /fusiform 22191
gyri
Right
occipital 22916
gyri/fusiform gyri
Left posterior insula/SII
3155
Right posterior insula/SII 2871
Sad and Left anterior cingulate 157
Pain
(BA 32)
(P<0.001) Right medial prefrontal 308
(BA9)
Left
1349
amygdala/hippocampus
Right
901
amygdala/hippocampus
Left
occipital/fusiform 26964
gyri
Right occipital/fusiform 20382
gyri
Right anterior insula
1948
Left thalamus
149
Left PAG
171

Mean z
value

Max z
value

Center location
(x,y,z)

3.44 ± 0.01
3.59 ± 0.01
3.47 ± 0.01
3.3 ± 0.01

3.72
4.33
3.86
3.44

(-44, -25, 47)
( -44, -25, 17)
(47, -29, 20)
(-10, 7, 39)

4.05 ± 0.01

4.39

(0, 48, 26)

4.22 ± 0.01

4.92

(46, 15, 19)

4.40 ± 0.01

5.46

(-26, -24, -10)

4.23 ± 0.01

5.95

(26, -18, -7)

4.32 ± 0.00

6.10

(-35, -79, -4)

4.48± 0.00

5.35

(41, -70, -3)

3.89± 0.01
3.42± 0.00
3.15±0.00

5.23
3.83
3.29

(-42,-28,18)
(52, -25,23)
(-6, 4, 39)

3.43±0.01

3.84

(5, 45, 30)

4.07±0.01

4.73

(-18, -29, -2)

3.98±0.00

4.28

(15, -29, -3)

4.89 ± 0.00

6.50

(-41, -70, -10)

4.23± 0.00

5.29

(41, -69, -7)

3.38 ± 0.00
3.43± 0.08
4.00 ± 0.01

3.57
3.43
4.19

(34, 21, 14)
(-17, -18,8)
(-5, -28, -8)

Table 2.1 Brain regions activated during three experimental trials.

48

2.3.2.2. The Interaction of Sadness and Pain
Brain regions associated with the interaction of sadness and pain were revealed by a groupwise whole brain t-test contrast analysis (Sad and Pain – Sad Only – Pain Only). Positive values
in the active clusters revealed by this contrast analysis indicated significantly greater brain
activation in response to painful stimuli presented with sad pictures than to either painful stimuli
or sad pictures presented alone. At an alpha level of 0.005, corrected for the whole brain,
significant activation was observed in bilateral subgenual cingulate cortex (sACC), contralateral
(left)

posterior

periaqueductal

insula/secondary
grey

(PAG),

somatosensory

bilateral

cortex

(pIn/SII),

amygdala/hippocampal

contralateral

complex

and

(left)

bilateral

occipital/fusiform gyri (Figure 2.8 and Table 2.2). The contrast analysis (Sad and Pain – Sad
Only) revealed that, as in the Pain Only condition, bilateral pIn/SII and contralateral (left) ACC
were preferentially active during acute painful stimuli (Figure 2.9 and Table 2.2).

49

A

B

Z

≥5

0
C

D

≤ -5

Figure 2.8 Brain activation as a result of the contrast analysis (Sad and Pain – Sad Only – Pain
Only). All images are in the axial plane. Green crosshairs centered on region of interest. Left
hemisphere shown on left. A: bilateral subgenual anterior cingulate cortex; B: left posterior
insular/secondary somatosensory cortices (pIn/SII); C: left periaqueductal grey (PAG); D:
bilateral amygdala/hippocampal complex. Alpha value, corrected for the whole brain, is 0.005.
Z
A

B

C

≥5

0

≤ -5

Figure 2.9 Brain activation as a result of the contrast analysis of (Sad and Pain – Sad Only).
Images are in the parasaggital plane. Green crosshairs centered on region of interest. A: left
anterior cingulate cortex (BA 32); B: left posterior insular/secondary somatosensory cortices
(pIn/SII); C: right pIn/SII. Alpha value, corrected for the whole brain, is 0.001.

50

Condition

Cluster location

Left posterior
insula/SII
Left/Right subgenual
anterior cingulate
cortex
Left amygdala/
hippocampal complex
Right amygdala/
hippocampal complex
Left periaqueductal
grey
Left occipital/fusiform
gyri
Right occipital/
fusiform gyri
Left posterior
Sad and
insula/SII
Pain – Sad
Right posterior
Only
insula/SII
(P ≤ 0.001)
Left anterior cingulate
cortex
Table 2.2 Brain activation as results of
Only) and (Sad and Pain – Sad Only).
Sad and
Pain – Sad
Only – Pain
Only
(P ≤ 0.005)

Active
volume
3
(mm )
666

Mean z
value

Max z
value

Center location
(x,y,z)

3.52 ± 0.01

3.97

(-38, -22, 19)

94

2.94 ± 0.01

3.14

(-6, 25, -8)

106

3.37 ± 0.01

3.52

(-19, -12, -14)

476

3.41 ± 0.01

3.64

(19, -17, -11)

414

3.03 ± 0.01

3.49

(-6, -30, -11)

13191

3.62 ± 0.01

4.10

(-40, -71, -4)

13916

3.48± 0.00

3.85

(41, -70, -3)

2516

3.6 ± 0.005

4.34

(-40, -23, 18)

835

3.5 ± 0.006

3.87

(38, -15, 13)

162

3.35 ± 0.005 3.44

(-11, 2, 39)

contrast analyses of (Sad and Pain – Sad Only – Pain

2.4 Discussion
Previous behavioral studies have reported that emotional states can alter the perception of
pain, but the neural mechanisms that account for this phenomenon have not yet been elucidated.
In this study, we replicated this finding, observing that a sad state induced by showing
emotionally evocative pictures significantly increases subjective pain intensity. Further, we
found that neural activation associated with increased pain perception is significantly increased
in a region of the contralateral parietal operculum that includes posterior insular/secondary
somatosensory cortex (pIn/SII). This region is known to be an important area in the complex of
brain regions commonly found to be activated during the experience of pain, often referred to as

51

the “pain matrix” (Peyron et al., 2000; Apkarian et al., 2005). In addition to pIn/SII, three other
brain areas showed significant activation associated with increased pain perception during
sadness: the subgenual anterior cingulate cortex, amygdala/hippocampal complex, the occipital
gyri. These cortical areas are likely to be associated with emotional processing of sad pictures,
and it is possible that activation in these brain regions reflects increased sadness during the
presentation of sad pictures and pain stimulation. However, the behavioral data do not show
significant differences in the ratings of sadness during Sad Only and Sad and Pain conditions.
We therefore suggest that the activation in subgenual anterior cingulate cortex and
amygdala/hippocampal complex, together with that in the midbrain periaqueductal grey (PAG),
may be involved in the modulation of pain perception through the descending pain transmission
pathway.
2.4.1 Increased Activation in Posterior Insular and Somatosensory Cortices
The secondary somatosensory cortex occupies a large area in the parietal operculum. The
exact anatomical border between the granular cortex of the inner part of the operculum and that
of the posterior insular cortex cannot be delineated using MRI, and the majority of pain
neuroimaging studies do not differentiate the secondary somatosensory area from the
anatomically contiguous posterior insular cortex (Peyron et al., 2000; Apkarian et al., 2005).
Thus, as in other fMRI reports, to describe the activation evoked in this cortical region by acute
painful stimuli, this study uses a term that includes both regions, here called the “posterior
insular/secondary somatosensory cortex” (pIn/SII).
The pIn/SII is thought to be important for processing the sensory-discriminative dimension of
pain (Treede et al., 1999). Evidence for this is based primarily on PET studies in which
perceived pain intensity is correlated with regional cerebral blood flow in the pIn/SII (Coghill et

52

al., 1999), and fMRI studies in which participants’ ratings of the sensory-discriminative
dimension of mechanical pain are correlated with BOLD signal in this region (Maihöfner et al.,
2006). In the present study, we found bilateral activation in pIn/SII when painful shocks were
administered to the right index finger. In addition, activation in the left pIn/SII, contralateral to
the stimulation site, was significantly greater in both BOLD signal intensity and area of
activation when painful shocks were administered concurrent with sad stimuli compared to when
painful shocks were administered alone, even though the intensity of pain stimuli was identical
in the two conditions. We suggest that the increased activation in the left pIn/SII reflected the
increased behavioral ratings of subjective pain intensity. Further, based on what is known about
the anatomical circuitry of this region, we believe it is likely that the increased activation is
primarily due to the posterior insular cortex. In the Old World macaque monkey, axons of
neurons in nociceptive specific lamina I of the spinal cord project to the posterior part of the
ventral medial thalamic nucleus which in turn projects to the dorsal margin of posterior insular
cortex (Craig, 1995). In this study painful stimuli were applied to the right side of the body, and
it is therefore not possible to separate the potentially unique roles of contralateral and ipsilateral
pIn/SII. It may be that left pIn/SII is involved in altered pain perception regardless of whether it
is ispilateral or contralateral to the stimulation site.
2.4.2 Role of Anterior Cingulate Cortex
The anterior cingulate cortex (ACC) is an important brain region for processing the affective
component of pain (Rainville et al., 1997; Hsieh et al., 1999; Tölle et al., 1999). A 1997 PET
study in which hypnotic suggestions were used to alter the unpleasantness of noxious heat
stimuli without changing the perceived intensity demonstrated that regional cerebral blood flow
changes within the ACC were positively correlated with ratings of pain unpleasantness (Rainville

53

et al., 1997). Further evidence for the role of the ACC in processing the affective dimension of
pain is seen is a recent fMRI study examining BOLD responses to electrical pain-inducing
stimuli while participants viewed randomly presented pictures of sad, happy or neutral faces.
Although neither pain nor emotional ratings were collected until after the scanning session, the
researchers reported that postscan pain ratings were higher during presentation of sad faces than
during happy or neutral faces. Analysis of the imaging data in that study found that activation of
ACC was significantly greater during presentation of sad faces (Yoshino et al., 2010). Consistent
with other fMRI studies of pain processing (see reviews in Peyron et al., 2000; Apkarian et al.,
2005), the present study observed activation in ACC contralateral to the body site of painful
stimulation. However, we did not find that activation of ACC during the concurrent
administration of painful stimuli and sad pictures was significantly increased compared to
administration of painful stimuli alone. This difference in results between our study and that of
Yoshino et al. is perhaps best explained by a difference in experimental design. Unlike Yoshino
et al., we did not use emotional facial expression to induce sadness, but instead chose stimuli that
were more general and had been shown in pilot studies to reliably induce a sad feeling state.
Several fMRI studies have demonstrated that empathy for pain involves affective components of
pain and is correlated with activation of ACC (Singer et al., 2004; Jackson et al., 2006). Perhaps
presentation of sad faces in the Yoshino et al. study induced empathy in the participants, thereby
tapping into the affective dimension of pain and causing increased activation of ACC. The
activation pattern in the current study most likely directly influenced the sensory dimension of
pain processing and increased activation in the pIn/SII without exacerbating the affective
dimension or activation in the ACC. This neural activation pattern is consistent with our

54

behavioral findings that the ratings of sadness during the presentation of sad pictures were not
significantly different between conditions with and without painful stimuli.
2.4.3 Sadness and the Descending Pain Pathway
In addition to pIn/SII, brain areas located in subgenual anterior cingulate cortex, amygdala,
and the occipital gyri showed significant activation in the contrast analysis using the equation of
Sad and Pain – Sad Only – Pain Only, indicating more activity during the Sad and Pain
condition compared to either the Sad Only or the Pain Only conditions. Each of these three areas
has been reported to be associated with emotional processing. In a meta-analysis of 55 PET and
fMRI studies on emotional activation, 46% of sad induction studies consistently reported
activation of the subgenual anterior cingulate cortex (sACC) (Phan et al., 2002). Although
Reiman and colleagues (1997) found sACC activity to recall-generated but not film-induced
sadness, Phan et al. (2002) found that sadness induced by recall or auditory or visual stimuli was
associated with the activation of sACC. Activation of sACC may not always be observed when
the targeted emotional state (i.e., sadness) does not achieve a desired intensity. This can happen,
for example, when mood induction is acoompanied by additional cognitive tasks. Interestingly,
in resting state studies, hypometabolism has been found in the sACC of patients with clinical
depression, and activation of sACC increases when patients respond to pharmacologic treatment
(Mayberg et al., 1999; Mayberg et al., 2000). Further evidence of an association between sACC
activation and sadness is supplied by Liotti et al. (2000) who note increased sACC activation if
subjects are scanned after achieving a predetermined level of emotion intensity.
Many studies have found an association between amygdala activation and fear-related
emotions, including 30 of the 55 neuroimaging studies analyzed by Phan et al. (2002). In
addition, PET studies have found enhanced activation in amygdala with increasing intensity of

55

sad facial expression (Blair et al., 1999), and in response to viewing both pleasant and aversive
pictures (Hamann et al., 1999). Thus, the human amygdala may play a general role in processing
salient emotional stimuli, regardless of the valence.
Finally, the occipital cortex is commonly activated by a wide range of visual emotion tasks,
including viewing emotionally-evocative pictures and films, and faces with emotional
expressions (Lang et al., 1998; Lane et al., 1999; Morris et al., 1996; Beauregard et al., 1998).
Although it is possible that increased activation in the occipital cortex as well as in the sACC and
amygdala reflects increased sadness when pain stimuli are delivered to participants in a sad
emotional state, the behavioral data do not show increased ratings of sadness during the Sad and
Pain condition compared to the Sad Only condition. Instead, we suggest that, at least in the
sACC and amygdala, increased activation may indicate modulation of pain perception through
known pain transmission pathways. The medial (midline and intralaminar) thalamic nuclei
project directly to both the anterior cingulate cortex and amygdala, which in turn each project to
the nociception-regulating center of periaqueductal gray (PAG) (Vogt and Sikes, 2000; Sewards
and Sewards, 2002). Thus, the projections from the PAG to the thalamus, and from the PAG to
the insular/somatosensory cortices, provide pathways to integrate sad and painful information
within the brain regions of the “pain matrix”.
In this study, the PAG, like the sACC and amygdala, was significantly more active during the
Sad and Pain condition compared to either the Sad Only or Pain Only condition. The PAG can
either inhibit or facilitate pain processing (Porreca et al., 2002). In the present study, the PAG
was likely involved in the facilitation of pain processing, since increased BOLD signal in the
PAG occurred during conditions in which participants rated painful shocks accompanied with
sadness as more intense than painful stimuli delivered alone. Top-down pain facilitation has been

56

examined in numerous animal studies. For example, chemical or electrical stimuli, applied to the
rat anterior cingulate cortex (ACC) (which sends projections to the PAG), enhances the tail-flick
reflex (Calejesan et al., 2000). As in humans, the PAG in rats sends projections to the rostral
ventral medulla (RVM). The ON-cells in the rat PAG increase their firing rates just before a tailflick response to radiation heat (Fields et al., 2000). Thus, the ACC can indirectly communicate
with the RVM. The descending facilitation mediated by the ACC can be blocked by intra-RVM
microinjection of an antagonist of AMPA and kainate receptors (Calejesan et al., 2000), further
indicating the central role played by the PAG in the descending facilitation of noxious stimuli.
The amygdala also likely contributes to descending facilitation of pain processing. The
primate amygdala sends projections to the PAG and receives substantial input from temporal
visual-association areas (Aggleton et al., 1980). There are also robust anatomical connections
between the sACC and amygdala (Anand et al., 2005; Mobascher et al., 2009; Johansen-Berg et
al., 2008). Visual information with negative valence could heighten activity in the amygdala, and
thereby indirectly influence the activation of PAG. Animal studies suggest that activation in the
pain facilitation pathway appears to depend on an intact amygdala. Lesions of the central
amygdala in rats eliminate shock-induced hyperalgesia as measured by a decrease in vocalization
thresholds to shock (Crown et al., 2000). In the present study, the bilateral amygdala showed
significant activation during presentation of sad pictures, both alone and concurrent with pain
stimulation. Thus, it is possible that activation in both the amygdala and the PAG could reflect
the influence of sadness on pain processing. The anatomical projection from occipital lobe to the
amygdala, and from amygdala to PAG provide a pathway by which emotional information
processed in the occipital lobe and amygdala could modulate nociceptive signals, possibly
enabling the processing of salient stimuli.

57

2.4.4 Conclusion
In conclusion, sadness enhanced subjective pain perception, which was accompanied by
increased activation within posterior insular/secondary somatosensory cortices (pIn/SII).
Moreover, brain regions, including sACC and amygdala, together with the PAG, showed
increased activation when painful stimuli were administered to study participants experiencing a
sad emotional state. Based on these results, we suggest that increased activation in the pIn/SII
could be part of a top-down facilitation of pain processing via neural circuits that include the
ACC, amygdala, occipital lobe and PAG. These findings may be relevant to the comorbidity of
pain and depression, potentially providing clues to why pain-free persons with pre-existing major
depression are at increased risk for developing symptoms of chronic pain conditions (Breslau et
al., 2003).

58

Chapter 3
Functional Connectivity of Brain Networks for Pain Processing and for Interaction of Pain
and Sadness
3.1 Introduction
Pain perception can be dramatically affected by one’s emotional state. For example, negative
emotions, such as unpleasant or depressive mood, can enhance pain intensity and decrease pain
tolerance (Meagher et al., 2001; Tang et al., 2008). However, the neural networks underlying
interaction of pain and emotional state are not well understood. Functional imaging techniques,
such as positron emission tomography (PET) and functional magnetic resonance imaging (fMRI),
have been used to investigate human brain activation during either pain processing or emotional
experience (see reviews in Apkarian et al., 2005 and Dalgleish, 2004). These neuroimaging
studies consistently identify a set of brain regions that are active during pain processing,
including primary/secondary somatosensory, insular, anterior cingulate and prefrontal cortices,
and the thalamus. Brain regions associated with emotional experience include the anterior
cingulate and prefrontal cortices, and the amygdala. Of note, the prefrontal, insular and cingulate
cortices, are activated by either pain stimuli or self-generated emotions (Apkarian, 2005; Ressler
and Mayberg, 2007; Damasio et al., 2000). In addition, some studies suggest that these brain
regions may be involved in the interaction of pain and emotional processing. For example, when
painful thermal stimuli are administered to the right arm of major depressive patients, the
prefrontal and insular cortical regions show increased activation, compared to healthy subjects
(Bӓr et al., 2007). In another study, participants were more sensitive to painful stimuli when they
viewed pictures of sad faces than neutral or happy faces. In these individuals, the anterior

59

cingulate cortex and amygdala showed increases in activation during increased pain sensitivity
(Yoshino et al., 2010).
In Chapter 2, we reported that subjective perception of pain intensity increases when
individuals experience a sad mood as a consequence of viewing sad pictures, and that this
behavior is accompanied by increased activation in the posterior insular/secondary
somatosensory cortex (pIn/SII). In addition, we found increased activation in some emotionrelated brain regions such as the subgenual cingulate cortex and the amygdala/hippocampal
complex. These observations led us to wonder how discrete brain regions work together in a
network to increase the perception of pain during sadness. To answer this question, it was
important to identify the functional connectivity among brain regions that are active during the
processing of either pain or emotion. We therefore included the insular cortex, cingulate cortex,
somatosensory cortex, and amygdala in our investigation of the network that could underlie the
altered pain perception during emotional states.
Traditional task-driven fMRI studies detect increased regional cerebral blood flow (rCBF) in
response to specific experimental tasks. For example, in the fMRI study of Chapter 2, we
designed three trials to investigate the brain activation patterns associated with painful and sad
stimuli. In recent years, researchers have become increasingly interested in investigating the
brain activity during rest. The resting-state fMRI detects the spontaneous blood-oxygen-leveldependent (BOLD) signal which is usually minimized by averaging in the traditional task-driven
fMRI studies. It is assumed that if brain regions participate in a particular common function,
their spontaneous BOLD signals during rest should correlate with one another (see reviews in
Fox and Raichle, 2007).

60

The aim of this study was to use resting-state fMRI to examine the functional connectivity
within neural networks implicated in the interaction of pain and sadness. We used brain regions
identified in Chapter 2 as the seed regions to explore the intrinsic activation patterns in the brain
network associated with pain at rest. These regions, the insular cortex, anterior cingulate cortex,
somatosensory cortex, and amygdala, were also used as candidates in a path analysis to identify
functional connectivity maps of brain networks. We predicted that these four candidate brain
regions would be functionally correlated at rest, and that their functional connectivity would be
altered when painful conditions were accompanied by and emotion-inducing stimuli.
3.2 Methods
3.2.1 Subjects
Sixteen healthy, right-handed women ages 18 – 33 years (mean 21.9) participated in the
study after providing written informed consent approved by Michigan State University’s
institutional review board.
3.2.2 Image Acquisition
®

Images were acquired on a GE 3T Signa HDx MR scanner (GE Healthcare, Waukesha, WI)
with an 8-channel head coil. A high-order shimming procedure was conducted to improve
magnetic field homogeneity (Kim et al., 2002). During the 7-minute resting-state scan, subjects
*

were instructed to relax, stay awake, and to keep their eyes closed. T2 -weighted echo planar
images were collected continuously with the following parameters: 36 interleaved 3-mm axial
°

slices, TR = 2.5 s, TE = 27.7 ms, flip angel = 80 , FOV = 22 cm, matrix size = 64 × 64. The first
four time points were discarded during data analysis to ensure that analyzed images were
collected after the magnet had reached a steady state. To aid in spatially locating resting-state

61

fMRI data, high-resolution T1-weighted inversion recovery fast spoiled gradient recalled echo
(IR FSPGR) images were acquired with the following parameters: 180 sagittal 1-mm slices, TE
°

= 3.8 ms, TR = 8.6 ms, flip angle = 8 , FOV =25.6 mm and matrix size = 256 x 256.
3.2.3 fMRI Data Preprocessing and Analysis
All images were analyzed using the Analysis of Functional NeuroImages (AFNI) software
package (Cox, 1996). Using the last functional image as a reference, images for each subject
were corrected for head movements in three translational and three rotational directions. Images
were then smoothed with a 4-mm FWHM Gaussian kernel to reduce spatial noise. The effects of
head movements, baseline, and system drifting were removed from the original signal through
multiple linear regression analysis. The cleaned echo planar imaging (EPI) signals in each time
series were then resampled to a voxel size of 3× 3 × 3 mm and converted to the Talairach and
Tournoux stereotaxic coordinates (Talairach and Tournoux, 1988) for further analysis.

62

M1

S1

SMA
S2
ACC

PCC

PPC

INSULA
MDvc
PF

VMpo

VPL

HT
AMYG

PAG
PB

Figure 3.1 Schematic of ascending pathways, subcoritical structures and cortical structures
involved in processing pain (Price, 2000). It is as same as Figure 1.3 in Chapter 1.
Based upon the known anatomical connections of subcortical and cortical pain processing
regions (Figure 3.1, Price, 2000) and our findings from Chapter 2, a hypothetical model for pain
processing was constructed. The following areas were included in the model, because they were
observed to be significantly active during pain processing: (1) left primary somatosensory cortex
(SI), (2) left posterior insula/secondary somatosensory cortex (pIn/SII), (3) left anterior cingulate

63

gyrus/BA 24 (ACC), (4) left posterior cingulate gyrus (PCC), (5) left thalamus (Tha) and (6) left
periaqueductal gyrus (PAG).
In addition, a hypothetical model for the interaction of sadness and pain was constructed
based upon the significant activation observed in the brain regions when pain perception was
altered by sadness. The model included all of the brain areas in the pain processing model as well
as: (1) bilateral medial prefrontal cortex (mPFC), (2) right inferior frontal gyrus (IFG), (3) right
anterior insula (aIn), (4) bilateral amygdala/hippocampal complex (AMY/Hip), (5) left subgenual
cingulate gyrus/BA 25 (sACC), (6) left parietal lobule/BA7 (PPC) and (7)

bilateral

occipital/fusiform gyri (OCP). The ROIs of posterior cingulate gyrus and parietal lobule/BA7
were generated by using atlases in the AFNI software. All other ROIs were adopted from the
imaging results of Chapter 2 (See Talairach coordinates of the centers of these ROIs in Table
3.1).
Central coordinates of ROIs
ROIs
(Right+/Left-, Anterior+/Posterior-, Superior+/Inferior-)
SI
(-44, -25, 47)
pIn/SII
(-45, -28, 18)
ACC
(-10, 7, 39)
PCC
(-5, -50, 18)
Tha
(-17, -18, 8)
PAG
(-5, -28, -8)
mPFC
(0, 48, 26)
IFG
(46, 15, 19)
aIn
(34, 21, 14)
AMY/Hip
(-26, -24, -10)
sACC
(-6, 25, -8)
PPC
(-27, -58, 49)
OCP
(-35, -79, -4)
Table 3.1 Regions of Interest (ROIs)

3

size (mm )
479
3155
162
6998
149
171
550
1979
1948
2555
94
15255
22191

3.2.4 Simple Correlation analysis and group ICA analysis
Simple correlation analysis used pre-defined seed regions (see below) to detect spatial
patterns of spontaneous fluctuations in the pain network during the resting state. Correlation

64

coefficients were calculated between a given seed region and each voxel in the entire brain (Fox
et al., 2005; Greicius et al., 2003). A separate independent component analysis (ICA) was used to
analyze the resting-state fMRI data without pre-defined seed regions. ICA assumes that the
BOLD signal is composed of linearly-additive independent components. Statistically, these
components are maximally independent from one another and each component is associated with
a spatial brain activity map. The group-level independent maps can be generated using group
independent component analysis (gICA) algorithm (Calhoun et al., 2004).
See regions for the correlation analysis were the posterior insular/secondary somatosensory
cortex (pIn/SII), anterior cingulate cortex (ACC) and medial prefrontal cortex (mPFC) (see Table
3.1). These three seed regions were important in both pain and emotion networks. In Chapter 2,
we observed that the pIn/SII responds to painful shocks and has significantly increased BOLD
signals when sad pictures were presented concomitant with painful shock. The ACC was
activated by either pain or sad stimuli. The mPFC was activated by the presentation of sad
pictures. To obtain a correlation map, averaged BOLD signals of a seed region during the
resting-state scan were first extracted. The corrections between the average time course at a seed
region and the time courses of all voxels in the brain were then performed. In group analyses, to
normalize the distribution, the R values were first converted to Z values through Fisher's Z
transformation. T tests were then performed on the Z values against random chance. A statistical
threshold of P < 0.001 (uncorrected for the whole brain) was used to determine which brain areas
were significantly correlated with the seed region across subjects.
We also used group ICA (Calhoun, 2001) to examine spatial patterns during rest. In this
analysis, the fMRI dataset for each subject was pre-processed to remove head movement and
global system trend described earlier (e.g., time-shifting of slices, motion correction, smoothing

65

and normalizing BOLD signals). We then applied principle component analysis (PCA) on
individual cleaned datasets to represent most of the variance of the fMRI within a lower
dimensionality (McKeown et al., 1998). The dataset for each subject had fewer than 25 sources
of variance; therefore, the software package of 3dpc in AFNI (Cox, 1996) was used to perform
PCA with a dimensionality reduction of 25. Because the results were insensitive to the reduction
parameter, the dimension of 25 is likely to be a good choice for decreasing computational load in.
Preprocessed datasets for all sixteen subjects were concatenated into one dataset and the
dimensionality of the aggregated data was then reduced to 20 by using 3dpc. Because there were
fewer than 20 ROIs, we chose 20 to estimate the number of sources present from the dataset to
provide a tradeoff between preserving much of information in the dataset and making the
calculation and interpretation less intensive. After the dimensionality reduction, the software
package

3dICA.R

in

AFNI

was

used

to

run

the

ICA

analysis

(http://afni.nimh.nih.gov/sscc/gangc/ica.html). The analysis produced representative time courses
of 20 independent components and mixing coefficients associated with these time courses. For
each component, the mixing coefficient of an individual voxel showed the weight of
corresponding time course component in the original BOLD signal. The group maps of 20
independent components generated from ICA were thresholded by using a Z-threshold criterion
(P < 0.001, Z = 2.36); ICA components were treated as random variables, and a one-sample t-test
was performed with the null hypothesis of zero magnitude.
3.2.5 Functional Connectivity Models
Simple correlation analyses demonstrate how one seed region functionally connects to other
brain regions at rest.

If we want to know correlations among multiple brain regions, the

simultaneous functional connectivity among them can be revealed using path analysis. Path

66

analysis is a causal modeling method to explore correlations within a defined network. However,
path coefficients generated from path analysis are not correlation coefficients. Theoretically, path
coefficients are standardized versions of linear regression weights which are used to determine
the relationship among variables in a mulitivariate system.
Model construction: Two hypothetical models were constructed to investigate brain networks
for pain processing and the interaction of pain and sadness (see Figures 3.2 and 3.3). The
hypothetical model in Figure 3.3 was simplified to include only ROIs significantly correlated
with others (see Table 3.3, the inter-regional Pearson’s correlations). For example, the brain
activation pattern during rest (Figure 3.5, 3.6 and 3.7 B) suggested strong correlations between
the anterior and posterior portions of cingulate cortex, as well as between the anterior insular
cortex and anterior cingulate cortex. Thus, the ROIs of ACC, PCC and aIn, and the anatomical
connections of these ROIs (Figure 3.1) were included in the hypothetical model. Uncertain
directionality of effect between two regions in the hypothetical model was taken care by: (1)
trying alternative models with different directionalities; (2) including reciprocal effects; or (3)
including all possible paths.
Model analysis: Path analysis was used to explore correlations within the defined networks
during both resting and task-driven states. For each subject’s dataset, masks of ROIs were
created. Within each mask, preprocessing steps were performed across all brain voxels as
detailed in section 3.2.3 to remove head movement and system drift. Preprocessed BOLD signals
were normalized by dividing the value of each voxel by the mean value of the whole brain. The
normalized time series of each mask were then averaged and extracted. The extracted time series
were used as inputs for path analysis using the program Amos 17.0 (SPSS Inc.).

67

Path analysis was applied as a confirmatory test to explore the correlations among ROIs
within the hypothetical models. All variables representing BOLD signals of ROIs were
observable (endogenous) variables; the analysis did not involve any latent (exogenous) variables.
The analysis estimated parameters of path coefficients and the goodness of fit which was used to
reject, modify or accept the model. Modification of the model was made one path at a time until
the overall goodness of fit was acceptable for the whole model. To achieve an acceptable
goodness of fit, a measure of discrepancy is minimized between the observed interregional
correlation matrix and the correlation matrix predicted by the model. The measures of this
discrepancy are asymptotically distributed as Chi Square. If observed discrepancy has a small
probability under this distribution, we can fairly sure that the model adequately fits the observed
data. If a small change in a path leads to a large change in other connections, this path would be
rejected because of co-linearity.
Model interpretation: Path analysis was based on the hypothesis that brain regions involved
in a functional brain network should have strongly inter-correlated activity patterns. For example,
if the path coefficient connecting region A to region B indicates that region A increases by one
standard deviation from its mean, then region B would be expected to change by the amount of
the path coefficient standard deviation from its mean, while all other relevant regional
connections were held constant. A positive path coefficient would indicate increased activation
in region A, leading to proportionately increased activity in region B.

68

SI

ACC

pIn/SII
PCC
Tha

PAG
Figure 3.2 A hypothetical model used to investigate the functional connectivity in the supraspinal
network involved in pain processing. SI, primary somatosensory cortex; pIn/SII, posterior
insular/secondary somatosensory cortex; PCC, posterior cingulate cortex (BA 30); ACC, anterior
cingulate cortex (BA 24); Tha, ventral lateral posterior nucleus of thalamus; PAG,
periaqueductal gyrus.

69

PPC

SI
IFG
pIn/SII

ACC
mPFC

PCC
aIn
Tha

AMY/Hip

sACC
OCP
PAG
Figure 3.3 A hypothetical model used to investigate the interaction of pain and emotion. This
model combines the brain regions involved in both pain and emotional processing. SI, pIn/SII,
PPC, Tha, and PAG are brain regions mainly associated with pain processing. IFG, mPFC, and
OCP are involved in processing emotional information. The ACC, aIn and AMY/Hip ROIs (bold)
are the probable key nodes modulating the pain experience by emotional states.

3.3 Results
3.3.1 Resting-state Maps
During the resting state, the spontaneous BOLD signals of the seed region placed at the
posterior insular/secondary somatosensory cortex (pIn/SII) were significantly correlated with
signals in the following bilateral brain regions: an area in the inferior frontal junction which
included the anterior insular cortex; anterior cingulate cortex (BA24); primary sensory cortex
extending to the parietal lobule; and the thalamus (Figure 3.4). Another seed region was placed
in the anterior cingulate cortex (ACC), an area which may be important in integrating emotion
and pain. This ACC seed region was significantly correlated with the anterior insular cortex and

70

the thalamus (Figure 3.5). There was also a strong correlation between regions in the anterior and
posterior portions of the brain. A seed region placed in the medial prefrontal cortex (mPFC) was
positively correlated with anterior regions (anterior insular and anterior cingulate cortices) and a
posterior region (posterior cingulate cortex) (Figure 3.6).
Independent Components Analysis (ICA) was also used to confirm functional connectivity
without a priori definition of seed regions. Two of the 20 component maps obtained in this
analysis were potentially relevant to pain and emotional networks (Figure 3.7). Each independent
map represented the weight of a component in the normalized time series during the resting state.
th

The 10 component map (IC 10) included large areas within the somatomotor cortex, extending
to the posterior insular cortex. This suggests strong positive correlations between the
th

somatomotor system and the insular cortex in human brain. The 17

component map (IC 17)

indicated high levels of activity during rest in bilateral cingulate cortex and medial prefrontal
cortex. This spatial pattern suggests strong positive correlations between the anterior portion of
the cortex (including medial prefrontal and anterior cingulate cortices) and the posterior portion
(including posterior cingulate cortex), consistent with the well-known activation pattern at rest in
humans (Greicius et al., 2003).

71

y=13
2

3

1

ROI

-11

1
4

5

0.5

6

0

-35

-0.5

7

r

-1

Figure 3.4 Coronal views in standard Talairach stereotactic space of whole-brain interregional
functional connectivity during the resting state at a group-level correlation analysis. A predefined seed ROI was placed in the posterior insular/secondary somatosensory cortex, P < 0.001,
correlation coefficient r > 0.2. Positive correlations of spontaneous fluctuations between the
brain voxels and the ROI voxels are shown in yellow, orange and red. These regions included: 1,
inferior frontal junction/anterior insular cortex; 2, anterior cingulate gyrus (BA 24); 3, inferior
frontal and superior temporal gyri, anterior/middle insular cortex; 4, posterior insular/secondary
somatosensory cortices; 5. primary somatosensory cortex/parietal lobule; 6. thalamus; 7.
secondary somatosensory cortex/parietal lobule.

72

y=36
ROI

12

1
1

2

0.5
0

-12

-0.5

r

-1

Figure 3.5 Coronal views in standard Talairach stereotactic space of whole-brain inter-regional
functional connectivity during the resting state at a group-level correlation analysis. A predefined seed ROI was placed in the anterior cingulate cortex, P < 0.001, correlation coefficient r
> 0.2. Positive correlations of spontaneous fluctuations between the brain voxels and the ROI
voxels are shown in yellow, orange and red. These regions included: 1. insular cortex; 2.
thalamus.

73

-34 (L)
ROI

1

1

-7
2

3
0.5
0

20(R)
-0.5
-1
r

Figure 3.6 Sagittal views in standard Talairach stereotactic space of whole-brain interegional
functional connectivity during the resting state at a group-level correlation analysis. A predefined seed ROI was placed in the medial prefrontal cortex, P < 0.001, correlation coefficient r
> 0.2. Positive correlations of spontaneous fluctuations between the brain voxels and the ROI
voxels are shown in yellow, orange and red. These regions included: 1. anterior insular cortex; 2.
prefrontal and anterior cingulate cortices; 3. posterior cingulate cortex.

74

A. IC 10
y=-4

12
-28

6
0
-6

-52

-12
Z
B. IC 17
y = 20
y=20

-4

-28

th

th

Figure 3.7 A. The 10 independent component map and B. the 17 independent component map
(P<0.001, z = 2.06). The temporal patterns of regions in yellow/orange have positive component
weights.
75

3.3.2 Connectivity Model of Pain Network
Pearson correlation coefficiewnts among the regions of interest (ROIs) in the hypothetical
model (Figure 3.2) were calculated for both the resting state and during painful stimulation. The
six ROIs in the model of Figure 3.2 were significantly correlated with one another (Table 3.2 and
Table 3.3). Next, the path analysis calculated the causal relationships among these ROIs during
painful stimuliation. The model illustrated in Figure 3.2 was modified to reach acceptable
goodness-of-fit (Figure 3.8: χ 2 (11) = 35.45, p < 0.001; Figure 3.9: χ 2 (11) = 54.71, p < 0.001).
The paths included in the final result all had regression weights that were significantly greater
than zero at a p level of 0.001. The pain network during the resting state is depicted in Figure 3.8
with eight statistically significant paths: PAG→Tha; Tha → SI; Tha → ACC; Tha → pIn/SII;
SI→ pIn/SII; pIn/SII→ ACC; ACC → PCC; PCC→ ACC. During painful stimulation, the level
of functional connectivity increased in four of the paths: Tha →SI; Tha → pIn/SII; SI → pIn/SII;
and pIn/SII →ACC (See Figure 3.9, bold arrows). Interestingly, the path coefficient between
PAG and Tha changes from a postive value to a strongly negative value, suggesting that the PAG
has an inhibitory effect on the pain-perception network.
3.3.3 Connectivity Model for Interaction of Pain and Sadness
Path analysis was also used to analyze the hypothetical model depicted in Figure 3.3 during
the interaction of pain and sadness. The model was modified to reach acceptable goodness-offitness (Figure 3.10: χ 2 (35) = 126, p < 0.001). The final model presented in Figure 3.10
contains three sub-networks, one each for pain, emotion, and the interaction of the two. The
significant paths during painful stimulation (Figure 3.9) were also preserved in the emotion/pain
model (Figure 3.10): i.e., (1) Tha → SI → pIn/SII →ACC and (2) Tha → pIn/SII. Moreover, the
ACC node integrated efferent and afferent information from other subdivisions of cingulate

76

cortex, and the insular cortex, PAG, and amygdala to form several functional circuits: i.e., (1)
pIn/SII → aIn → ACC → PCC, (2) ACC → sACC → Tha →pIn/SII, (3) ACC → AMY/Hip,
and (4) ACC → PAG →pIn/SII. These functional circuits reveal that the ACC is in a key
functional postion to influence activity in brain regions that process pain perception. Recall from
Chapter 2 that the pIn/SII shows increased activation when painful stimuli are adminsterd when a
subject is in a sad mood, resulting in increased pain perception (Figure 2.8 in Chapter 2). In the
path analysis described here, this pIn/SII ROI receives positive inputs from SI, Tha and PAG, as
well as negative input from PPC. Through the node of PPC, there is a combination of positive
and negative path coefficients: (1) AMY/Hip → PCC →PPC; (2) ACC → PCC → PPC; and, (3)
sACC → PCC → PPC. Thus, the combined effect of these paths has positive influence in the
path of PPC → pIn/SII. It suggests the increased activation in AMY/Hip, ACC and sACC leads
to the increased activation. Similarly, the positive feedback loop from the pIn/SII → aIn → ACC
→ PCC → pIn/SII results in an increased activation in the pIn/SII. Finally, the subpath PAG →
pIn/SII, which also integrates the inputs from ACC and AMY/Hip, resulted in additional positive
influences on pIn/SII.

ACC
PAG

ACC
1
.220

PCC

.411

pIn/SII

.540

SI

.545

**
**

PAG

PCC

pIn/SII

SI

Tha

1
**

1

**

.203
0.057

**

0.115

.543

**

**
**

1
.657

**
**

1

**
1
.499
.250
.509
.335
.368
Table 3.2 Input data for analyzing the functional connectivity of the pain processing network
(Figure 3.2) during rest. Pearson correlations between the ROIs (** p < 0.01)

Tha

**

**

.573

77

ACC
PAG

ACC
1
-.520

PCC

.151

pIn/SII

.752

SI

.783

PAG
**

PCC

pIn/SII

SI

Tha

1

**

-.248

**

-.787

**

-.712

**

1

**

**

1

.382

**

**

.305

.91

**

1

**
**
**
**
**
1
.564
-.747
.172
.776
.717
Table 3.3 Input data for analyzing the functional connectivity of the pain processing network
(Figure 3.2) during painful stimulation. Pearson correlations between the ROIs (** p < 0.01)

Tha

ACC
sACC
aIn

ACC
1
.661
.453

sACC

**
**

AMY/Hip .579**
**
pIn/SII
.761
SI

.658

Tha

.704

PAG

.755

PCC

.775

**
**
**
**

pIn/SII

SI

Tha

1
.347
.592
.515
.534
.730
.558
.676

**

1

**

**

.319

**

**

.402
0.245

**
**

**

.286

**

**

.458

**

**

PPC

AMY/H
ip

aIn

**

.341

**

1
**

.791

**

.522

**

.822

**

.615

**

.615

**

**

1
**

.704

**

.754

**

.788

**

.660

**

1
**

.673

**

.520

**

.542

**

1
**

.567

**

.730

**

-.492
-.305
-.277
-.441
-.658
-.470
-.420
Table 3.4 (A) Input data for analyzing functional connectivity of Figure 3.3 when painful stimuli
were administered during sad mood. Pearson correlations between the ROIs (** p < 0.01)

PAG
ACC
sACC
aIn
AMY/Hip
pIn/SII
SI
Tha
PAG
1
**
PCC
.665
**
PPC
-.552

PCC

PPC

1
-0.253

1

Table 3.4 (B) Input data for analyzing functional connectivity of Figure 3.3 when painful stimuli
were administered during sad mood. Pearson correlations between the ROIs (** p < 0.01)

78

SI
0.687

0.353

0.368

pIn/SII
ACC

0.437
-0.535

0.394

PCC
0.166
Tha
0.25
PAG

Figure 3.8 Functional connectivity for the pain network during rest. Standardized path
coefficients, computed from data in Table 3.2, represent the combined effect of all connections
included in the path analysis. The final model is modified based on the hypothetical model in
Figure 3.2.
SI
0.717

ACC

0.916
0.717

pIn/SII

0.438
-0.343

PCC
0.727

0.471
Tha

-0.747
PAG

Figure 3.9 Functional connectivity for the pain network during acute painful stimulation.
Standardized path coefficients, computed from data in Table 3.3, represent the combined effect
of all connections included in the path analysis. Bold arrows indicate increased path coefficients
of functional connectivity during painful stimulation compared to resting state.

79

SI
0.69
PPC

0.4
0.96

ACC
0.27

-0.45

-0.22

0.55

-0.23
0.28

PCC
0.35

aIn

pIn/SII
0.36
0.38

0.44

0.72
0.73

AMY/Hip
Tha

sACC

0.62

0.39

0.59
0.81

PAG

0.36

Figure 3.10 Functional connectivity for interaction between painful and sad conditions.
Standardized path coefficients, computed from data in Table 3.4, represent the combined effect
of all connections included in the path analysis. The ROIs primarily associated with the
processing of pain perception, including SI, pIn/SII, PPC and Tha, are indicated by the dashed
double line boxes. The ROIs primarily associated with the processing of sad pictures, such as
ACC, sACC, aIn, PCC and AMY/Hip, are indicated by the dashed line boxes. PAG is a node that
receives inputs from emotional brain regions and is in a position to directly influence activation
in the somotosensory cortex. This final model includes statistically significant paths and is
modified based on the hypothetical model in Figure 3.3.
3.4 Discussion
To our knowledge, this is the first study to demonstrate functional connectivity among brain
regions within the pain matrix in which activity is modulated by sadness. The functional
connectivity of these brain regions during the resting and task-driven states provides insights into
understanding the altered pain experience influence by sadness.
3.4.1 Functional Network for Pain Perception

80

We set out to determine if correlations between the posterior insular/secondary
somatosensory cortex (pIn/SII) and other brain regions could be used to better understand the
brain activation pattern associated with pain processing, and, in the process, constructed a
hypothetical model of functional connectivity. The result of the study described in Chapter 2 is
that the posterior insular/secondary somatosensory cortex is significantly more active in response
to electrical pain shock than during a control condition. This is consistent with the discovery that
the SII, neighboring areas BA7b, and the posterior insula contain nociceptive neurons (Robinson
and Burton, 1980). These brain regions have consistently been found to encode the intensity of
nociceptive stimulation in both PET and fMRI neuroimaging studies (Peyron et al., 2002; Frot et
al., 2007). In the resting-state fMRI study described in this chapter, a correlation analysis
revealed that in both hemispheres the primary components of the acute pain processing network
(i.e., the anterior cingulate cortex (BA 24), SI, insular cortex and thalamus; see reviews in
Apkarain 2005) were positively correlated with our pre-defined seed region, the pIn/SII. An
independent component analysis conducted without the pre-defined seed region confirmed the
same pattern. Because the correlation patterns found during the resting state were also found
during task performance, these analyses demonstrate that, even in absence of external stimuli,
brain regions within the pain matrix maintain a level of functional connectivity. Comparing the
pattern of BOLD activity during rest with those during stimulus presentation, it is then possible
to quantitatively describe how external stimuli modulate the functional networks (Fox and
Raichle, 2007; Dhond et al., 2008).
We used the resting state activation pattern to construct a hypothetical neural network for
pain processing, so that we could employ path analysis to identify functional connections for
pain processing during both rest (Figure 3.8) and in response to acute pain (Figure 3.9). During

81

rest, significant functional connectivity was observed in a circuit beginning with the thalamus
and continuing to SI, SII, and the posterior insular and anterior cingulate cortices. The
anatomical connections of this functional pathway have been characterized in great detail in
several animal species, including humans (Price, 1999). Compared with the resting state, we
observed that path coefficients in this pathway increased during acute painful stimuli, seen
especially in the large path coefficient increases from thalamus to SI and to pIn/SII. The strong
positive interaction most likely means that increased activity in the thalamus directly leads to
increased activity in SI and pIn/SII.
We also observed increased functional connectivity from the posterior insular/secondary
somatosensory cortex to the anterior cingulate cortex during the response to pain. The ACC is
known to be an important region for processing the affective dimension of pain (Rainville, 1997).
The functional connection from insula to ACC suggests that the sensory dimension of pain may
be integrated with emotional and cognitive factors in the proposed network. This interpretation is
consistent with the elctrophysiological properties of nociceptive neurons in the ACC which
appear to monitor the overall state of the body with whole-body receptive fields (Vogt and Sikes,
2000).
3.4.2 Functional Network for Interaction of Pain and Sadness
We hypothesized that the pain network would directly interact with brain regions associated
with emotional processing. Previous reports have indicated that the ACC, amygdala, and anterior
insular cortex play important roles in processing both pain and emotion (Vogt, 2005; Rainville,
1997; Neugebauer et al., 2004; Craig, 2003). In this study, intrinsic connections between the
ACC and several brain regions, including the insular cortex, thalamus, and amygdala were
observed during rest. Using pre-established knowledge of the anatomical connnections among

82

those areas, as well as those with the PAG and prefrontal lobe, we created a model of functional
connectivity (Figure 3.10). In this proposed model, brain regions that process pain perception
primarily interact with subdivisions of cingulate cortex and the amygdala-hippocampal complex.
It is possible to decduce how these regions interact to produce a response to pain that is
influenced by emotional state.
The Periaqueductal Gray (PAG)
The model includes a direct functional connection from the ACC to the PAG. This is likely
to be the substrate through which the ACC can modulate activity in the PAG to either facilitate
or inhibit pain perception. The model also includes a direct connection from the amygdala to the
PAG. Studies have shown that the amygdala is activated when a person views unpleasant
pictures (Phan et al., 2002). In addition, individuals who have sustained sclerotic damage to the
amygdala do not show enhanced responses to emotional visual stimuli in visual cortex in fMRI
experiments (Vuilleumier et al., 2004). The pathway from the amygdala to the PAG, then, may
provide another substrate for the emotional context of the pain stimulus to subsequently affect
the PAG’s pain-gating function. Finally, the direct pathway from the PAG to the pIn/SII in the
model suggests that increased activity in the PAG can lead to increased activation in the pIn/SII.
It is possible that the PAG node integrates information from the ACC and amygdala, and
intensifies activation in the ascending pain pathway.
The Anterior and Posterior Cingulate Cortices
In our model shown in Figure 3.10, the amygdala and anterior cingulate cortex (BA24) are
functionally connected to the subgenual (sACC). The sACC is also functionally connected to
brain areas of the pain matrix through the thalamus. The sACC is known to be involved in
sadness and depression (Mayberg et al., 1999; Liotti et al., 2000). The model suggests that

83

information about the emotional context of pain may be integrated in the sACC. The sACC also
projects to the posterior cingulate cortex (PCC), consistent with our resting-state results showing
connections from the ACC, sACC and amygdala to the PCC, and may also integrate emotional
context with noxious information in PCC. A previous study from our laboratory showed that the
PCC is often activated by acute pain stimuli (Symonds et al., 2006). The bilateral connection
between the PCC and the parietal lobule/BA7 (PPC) suggests that information may be relayed
from the PCC to the pIn/SII, where it may influence activity in this primary node of the pain
matrix.
The Anterior Insular Cortex
In our model the anterior insular cortex (aIn) receives input from the pIn/SII, and projects to
the ACC. The aIn is thought to be involved in processing introspective feelings, including
sensory coding, body state assessment and autonomic regulation (Craig, 2009). Therefore, like
the sACC, it may also integrate information about pain with its emotional context.
The Prefrontal Cortex
The model in Figure 3.10 does not include prefrontal regions even though they were
originally proposed in the hypothetical model seen in Figure 3.3. This is because the inclusion of
prefrontal regions did not contribute to the validation of the functional connectivity model. It is
possible that there were few cognitive factors involved in the emotional modulation of pain
experience in the current study and/or that emotional modulation of pain processing is mainly
dependent upon activation in the cingulate cortex and amygdala.
3.4.3 Conclusion
The study described in Chapter 2 demonstrated that sadness can increase pain perception, and
that this is accompanied by significantly increased brain activation in the posterior

84

insular/secondary somatosensory cortices. The study described in this chapter reveals the
intrinsic functional connectivity among the main brain regions associated with pain perception at
rest. Specifically, the resting state connections between the thalamus and somatosensory cortices
show increased path coeffcients in response to pain stimuli. In addition, the results of a path
analysis revealed the direct interaction of pain and sadness in regions that process pain and
emotional information. Emotional context appears to influence pain perception through
functional connectivity among somatosensory, insular, posterior parietal and cingulate cortices,
amygdala and periaqueductal gray. Finally, the total effect of emotional context on pain matrix
areas through the projections to the posterior insular/ secondary somatosensory cortices is likely
to be communicated directly through the periaqueductal gray. These outcomes of path analysis
provide specific explanations for the interaction of pain perception and sadness and can now be
explicitly tested empirically.

85

Chapter 4
General Discussion
Emotion and pain are both complicated physiological and psychological constructs, and have
received extensive attention in neuropsychological research. Both are considered to be
multidimensional, comprising valence and sensory components. Clinical and laboratory studies
indicate that negative emotion, particularly sadness, can either increase pain perception or
decrease pain tolerance (Meagher et al., 2001; Tang et al., 2008). Depression and chronic pain
are often comorbid, providing a clinical impetus to understand the interaction between emotion
and pain at the neural level (Breslau et al., 2003).
A number of brain regions that respond to either pain or emotion have been identified in
recent years by neuroimaging studies (see reviews in Peyron et al., 2000; Dalgleish, 2004). Some
of these studies suggest that because the anterior cingulate cortex (ACC) is activated by both
pain and negative affect (Rainville et al., 2002; Vogt, 2005), it may be an important component
in the modulation of pain perception during different emotional states. However, it is not clear by
which neural circuits the ACC could be involved in the interaction between pain and sadness.
The aim of this dissertation is to test the hypothesis that sadness can influence subjective pain
perception, and to explore the functional connectivity of brain networks that are involved in
processing both pain perception and sadness. Specifically, the experiments described in Chapters
2 and 3: (1) identify human brain activation during perception of acute pain and how that activity
is modified during sadness; and, (2) construct the functional connectivity of neural networks
associated with pain processing as well as the interaction between pain and sadness.
4.1 Brain Regions Encoding both Pain and Emotions

86

Cortical and subcortical regions that process pain perception have been intensively
investigated in both healthy subjects and patients with clinical pain conditions. Across a variety
of experimental manipulations, researchers find that pain is not encoded in a singular “pain
center” of the brain, but is instead located within a network of somatosensory, limbic and
associative brain areas that receive parallel input from multiple nociceptive pathways (Peyron et
al., 2000; Apkarian et al., 2005). The main components in this brain network for pain processing
include primary and secondary somatosensory (SI/SII), insular, anterior cingulate (ACC) and
prefrontal cortices and thalamus. Increased regional blood flow changes to noxious stimuli are
consistently observed in SI/SII, insular and ACC cortices. The SI/SII and insular cortices are
thought to be related to the sensory-discriminative aspects of pain perception such as the
intensity or location of a painful stimulus (Coghill et al., 1999; Bushnell et al., 1999). The ACC
is observed to be activated in the majority of PET or fMRI studies, but does not seem to be
involved in encoding either stimulus intensity or location, and instead has been implicated in
affective aspects of pain processing (Rainville et al., 1997; Tölle et al., 1999). For example, the
ACC is particularly active when a painful stimulus is perceived as more distressing, regardless of
its intensity (Rainville, et al., 1997). Consistent with current neuroimaging, in this study we also
find significantly increased BOLD activity in response to acute painful electrical shocks in the
contralateral primary somatosensory (SI), bilateral posterior insular/secondary somatosensory
(pIn/SII), contralateral anterior cingulate (BA 32), and in the thalamus.
The neuroanatomy of emotion in the human brain has also been the focus of recent
neuroimaging studies. The prefrontal cortex, amygdala, and anterior cingulate cortex appear to
be key brain regions involved in the emotional experience (Dalgleish, 2004). In a meta-analysis
of emotion activation studies in PET and fMRI, Phan et al. (2002) observe: (1) the medial

87

prefrontal cortex generally is engaged in the representation of elementary emotional states such
as happiness, sadness, disgust and the mixture of these emotions; (2) the amygdala plays a
crucial role in both the perception of emotional cues and the production and emotional responses,
with some evidence suggesting that it is specifically associated with fear-related emotions; (3)
the subgenual anterior cingulate (BA 25) cortex is significantly associated with sadness; (4) the
anterior cingulate and insular cortices are involved in emotions either with cognitive demand or
induced by emotional recall/imagery; and, (5) both the occipital cortex and the amygdala are
activated by emotions induced by visual stimuli.
In this study, subjective sadness was successfully induced by visual stimuli selected from the
International Affective Picture System (IAPS). Our results, consistent with observations of Phan
et al. (2002), show that watching sad pictures significantly activates the brain regions located in
the medial frontal cortex (BA9), the right inferior frontal cortex (BA 45/44), the bilateral
amygdala/hippocampus, and the bilateral occipital/fusiform gyri.
Not surprising, when we simultaneously present painful and emotional stimuli, significant
activation in brain networks for both pain and emotions was found. In addition, significant
activation was observed in the periaqueductal gray (PAG). This midbrain region was activate
most likely reflects the pain modulation in the ascending pathway. As the PAG also receives
direct inputs from emotion-related brain regions such as the ACC and the amygdala, top-down
influences of emotion from the ACC and amygdala could be exerted on the PAG, thereby
modulating transmission of pain from PAG to the thalamus or to the pIn/SII.
4.2 Secondary Somatosensory and Posterior Insular Cortices
In Chapter 2, we report that in healthy right-handed female participants, pain perception is
altered by subjective emotion. Participants rated their pain perception significantly higher when

88

they received pain stimuli while concurrently watching sad pictures, compared to receiving
painful stimulation while watching neutral pictures. Interestingly, we also found that activation
in the contralateral posterior insular/secondary somatosensory cortex, a part of the brain network
for pain processing, was significantly greater when painful shocks were accompanied by sad
pictures than when painful stimuli were presented with emotionally neutral pictures.
The border of secondary somatosensory and insular cortices has not so far been clearly
delineated anatomically on MRI slices. Thus, the results reported in our fMRI study do not
separate the two regions, and instead consider them to be one functional region. The pIn/SII is
most likely associated with the sensory-discriminative aspects of pain processing, as these
regions contain nociceptive neurons. Many previous PET/fMRI studies have found reliable painrelated activity in this brain region (see review in Apkarian et al., 2005). For example, the pIn/SII
is functionally implicated in the discrimination of intensity of thermal stimuli (Craig et al., 2000;
Peyron et al., 1999). With electrodes implanted within SII and posterior insula in patients
referred for presurgical evaluation of epilepsy, Frot et al., (2007) reported that increasing thermal
intensities of CO2 laser stimulation enhanced evoked potentials in both insular and secondary
somatosensory cortices.
When pain is accompanied by sadness, we also observe that increased activation in other
brain regions such as dorsal anterior cingulate cortex (ACC), subgenual anterior cingulate cortex
(sACC), amygdala/hippocampal complex (AMY/Hip) and periaqueductal gray (PAG). Several
studies have shown that the activation of anterior cingulate cortex and amygdala/hippocampal
complex is associated with emotional processing. For example, subgenual cingulate cortex has
reduced metabolic during rest in patients with clinical depression (Mayberg et al., 1994). The
dorsal anterior cingulate cortex is involved in the affective dimension of pain processing

89

(Rainville et al., 1997). The amygdala is known to process emotional stimuli, including facial
expression (Phan et al., 2004). Interestingly, in the functional connectivity analysis described in
Chapter 3 we observed that the pIn/SII is functionally connected with these emotion-related
brain regions through one or multiple nodes.
When pain perception is increased by sadness, the connections between the pIn/SII and brain
regions that process emotions may lead to increased activity in the pIn/SII. The pIn/SII receives
projections from the ventroposterior lateral nuclei of thalamus and the PAG in the nociceptive
pathways, and this is one of the first cortical relay stations in the central processing of painful
stimuli (Treede et al., 2000). The active area of thalamus and the PAG during pain also interact
with brain regions that process emotions. For example, our functional connectivity model shows
that the thalamus receives positive functional inputs from subgenual anterior cingulate cortex
(sACC) and amygdala. The PAG also receives direct input from the anterior cingulate cortex and
amygdala. These functional connections suggest that the altered activation in the pIn/SII could be
due to activity in the thalamus and PAG nodes in the pain ascending pathway, which themselves
receive functional inputs from cortical regions associated with emotions, including the sACC,
ACC, and amygdala.
We also find that the pIn/SII receives positive functional inputs from two other nodes: the
posterior cingulate cortex (PCC), and the parietal lobule/BA 7 (PPC). The parietal lobule/BA 7
contains nociceptive neurons and activation in the pIn/SII as a result of painful stimulation can
extend into this brain region (Peryon et al., 2000). A previous fMRI study in our laboratory
shows that the posterior cingulate cortex is also activated by acute painful stimulation (Symonds
et al., 2006). In our model presented in Chapter 3, the posterior cingulate cortex receives
functional input from the sACC, ACC and amygdala. Thus, the posterior cingulate cortex is
90

possibly involved in integrating emotional contexts and noxious information. The effect of the
path from the posterior cingulate cortex (PCC) to the parietal lobule/BA 7 (PPC) would be to
increase activation in the pIn/SII.
4.3 Functional Connectivity Analysis of Pain-Emotion Interaction
To our knowledge, this is the first study demonstrating altered functional connectivity among
brain when pain perception is increased by a change in mood. Functional connectivity analysis
provides information not available using traditional fMRI contrast analysis methods. For
example, fMRI contrast analysis methods cannot distinguish between excitatory and inhibitory
activity, but instead rely upon regional changes in the oxygenation of hemoglobin due to
increased synaptic activity (Logothetis et al., 2001). On the other hand, functional connectivity
analysis is a promising strategy for determining whether rCBF changes are due to excitation or
inhibition. This analysis investigates statistical correlation of dynamic flow changes between
interconnected regions. Positive or negative path coefficients help to explain the activation of
excitation or inhibitation in a downstream brain region.
In Chapter 3, resting state fMRI reveals intrinsic correlations among brain regions that are
functionally connected. Based on the observed brain region activation in the task-driven fMRI
described in our experiments in Chapter 2, we place ‘seed regions’ in the posterior
insular/secondary somatosensory and dorsal anterior cingulate cortices to investigate the
interregional correlation patterns in the brain networks for pain and emotions, respectively,
during the resting state. The resulting resting-state map of simple correlation analysis indicates
that brain regions, including somatosensory, insular, anterior cingulate, and parietal cortices, are
intrinsically correlated. Without predefined seed regions, we also used the independent
component analysis (ICA) to determine the activation pattern of the pain network at rest. The

91

correlation map of ICA confirms the intrinsic correlations among somatosensory, insular,
cingulate and parietal cortices. These brain regions may therefore have similar functionality in
response to external stimuli.
Once the resting state maps was created, path analysis then showed that these regions are, in
fact, functionally connected with one another. The following functional paths was observed
during the rest: (1) projections from the thalamus to the primary somatosensory, dorsal anterior
cingulate and posterior insular/secondary somatosensory cortices; (2) a feedback loop starting
with primary somatosensory cortex, and going from there to posterior insular/secondary
somatosensory cortices, dorsal anterior cingulate cortex, posterior cingulate cortex, and finally
back to posterior insular/secondary somatosensory cortices; (3) bi-directional connections
between the dorsal anterior cingulate cortex and posterior cingulate cortex; and (4) a projection
from the preaqueductal gray to the thalamus. Interestingly, the path coeffcients of some paths
increased when painful electrical shocks are administered. Specifically, the path projecting from
the thalamus to the primary somatosensory cortex, continuing to the posterior insular/ secondary
somatosensory cortices (pIn/SII), and reaching the dorsal anterior cingulate cortex is
significantly increased. Similarly, the direct connection from the thalamus to the pIn/SII is also
significantly increased. These functional paths outline important relays and connections related
to pain processing. In the pain processing network, a higher positive path coefficient in the path
from the thalamus to the pIn/SII most likely means that increased activity in the thalamus
directly leads to increased activity in the pIn/SII.
In Chapter 3, we presented a model to explain how brain regions that are parts of the pain
matrix could directly interact with brain region associated with emotional processing. The
functional connectivity results described in Chapter 3 reveal that the dorsal anterior cingulate

92

cortex receives inputs from brain regions associated with sad emotions, including the amygdala,
subgenual cingulate cortex, and anterior insular cortex. Information about sad emotions can then
be relayed to pain processing areas via connections within the cingulate cortex, as well as
connections between the cingulate cortex and the thalamus, periaqueductal gray, parietal cortex,
and finally to the posterior insular/secondary somatosensory cortices (pIn/SII). Thus the
functional connections among these sadness- and pain-processing regions can lead to modulated
activity in the pIn/SII during sadness. This model and the functional connectivity among the pain
and emotional networks explains the observation in the Chapter 2 that increased pain perception
during sadness is accompanied by the increased brain activation in the pIn/SII.
4.4 Conclusions
In this study, we identified brain regions associated with either pain or sadness and
confirmed results from previous neuroimaging studies. We then demonstrated that sadness can
increase pain perception and that this behavioral result is accompanied by significantly increased
brain activation in the secondary somatosensory and posterior insular cortices. Finally, we
employed resting-state fMRI and functional connectivity analyses to reveal for the first time the
neural underpinnings of altered pain perception during sadness.
The connectivity analyses in this study reveal the intrinsic functional connectivity at rest
among brain regions associated with pain perception. The functional connections between the
thalamus and somatosensory cortex are significantly increased during painful stimulation. In
addition, the results of path analysis reveal how sadness can influence pain perception via
connectivity among somatosensory, insular, parietal and cingulate cortices, amygdala and
periaqueductal gray. The combined effect of activity in emotion-processing areas can then
influence the posterior insular/ secondary somatosensory cortices directly through the

93

periaqueductal gray. Thus, path analysis can provide a neural explanation for how pain
perception can be altered by sadness.
4.5 Limitations
There are some limitations in studies described in this dissertation. First, we only recruited
female subjects, so that the results from the study in Chapter 2 cannot be generalized to all men.
Second, large clusters covering the occipital/fusiform gyri and cerebellum were observed during
trials when sad pictures were presented. These large clusters cannot be easily separated into
independent small clusters at the alpha value we used. To obtain meaningful activation in other
brain regions such as posterior insular/secondary somatosensory cortices, we have to sacrifice
the analysis of these large clusters. Finally, path analysis is only used in a confirmatory way. It
would have been most useful if we already had a clear hypothesis to test. Our construction of the
hypothetical model mostly depended upon known anatomical connections.
4.6 Future Directions
In Chapter 2, we designed a successful fMRI experimental paradigm for investigating how
pain perception can be altered by sadness. We intend to extend this paradigm to investigate the
relationship between stress and pain, fear and pain, and finally, the relationship between negative
emotions and pain in chronic pain patients.
Using the resting-state fMRI paradigm described in Chapter 3, we can further compare the
spontaneous brain activation patterns in the pain and emotional networks among different groups,
such as chronic pain patients, depression patients and healthy control participants. One
possibility is that intrinsic functional connectivity of functional brain networks are altered in
different clinical conditions.

94

Finally, the functional connectivity models developed and described in Chapter 3 can be
applied to develop the novel therapies that target the brain circuitry in the cortex rather than the
periphery or spinal cord pathways. The models may be modified independently under each type
of chronic pain and the changes of functional connectivity may be used to predict the efficiency
of therapies.

95

APPENDICES

96

APPENDIX A

Programming scripts in AFNI for processing the imaging data of Chapter 2
#convert DICOM files from the MRI scanner into AFNI format
to3d -session Sec1 -prefix run01 -epan -time:zt 38 60 3000 seq+z './11548/3/SIM*'
to3d -session Sec1 -prefix run02 -epan -time:zt 38 60 3000 seq+z './11548/4/SIM*'
to3d -session Sec1 -prefix run03 -epan -time:zt 38 60 3000 seq+z './11548/5/SIM*'
to3d -session anatomy -prefix SPGR './11548/6/SIM*'
#!/usr/bin/env tcsh
echo "auto-generated by afni_proc.py, Fri Oct 3 11:39:49 2008"
echo "(version 1.28, Jan 28, 2008)"
# execute via : tcsh -x altpain.pre.GAMblur8 |& tee output.altpain.pre.GAMblur8
# -------------------------------------------------# script setup
# the user may specify a single subject to run with
if ( $#argv > 0 ) then
set subj = $argv[1]
else
set subj = AM5023
endif
# assign output directory name
set output_dir = $subj.results
# verify that the results directory does not yet exist
if ( -d $output_dir ) then
echo output dir "$subj.results" already exists
exit
endif
# set list of runs
set runs = (`count -digits 2 1 3`)
# create results directory
mkdir $output_dir
# create stimuli directory, and copy stim files
mkdir $output_dir/stimuli
cp ../../1DFiles/stim.1D $output_dir/stimuli

97

# ------------------------------------------------------# apply 3dTcat to copy input dsets to results dir, while removing the first 0 TRs
3dTcat -prefix $output_dir/pb00.$subj.r01.tcat run01+orig'[0..$]'
3dTcat -prefix $output_dir/pb00.$subj.r02.tcat run02+orig'[0..$]'
3dTcat -prefix $output_dir/pb00.$subj.r03.tcat run03+orig'[0..$]'
# and enter the results directory
cd $output_dir
# ------------------------------------------------------# run 3dToutcount and 3dTshift for each run
foreach run ( $runs )
3dToutcount -automask pb00.$subj.r$run.tcat+orig > outcount_r$run.1D
3dTshift -tzero 0 -quintic -prefix pb01.$subj.r$run.tshift
\
pb00.$subj.r$run.tcat+orig
end
# ------------------------------------------------------# align each dset to the base volume
foreach run ( $runs )
3dvolreg -verbose -zpad 1 -base pb01.$subj.r03.tshift+orig'[59]' \
-1Dfile dfile.r$run.1D -prefix pb02.$subj.r$run.volreg \
-cubic
\
pb01.$subj.r$run.tshift+orig
end
# make a single file of registration parameters
cat dfile.r??.1D > dfile.rall.1D

# ------------------------------------------------------# blur each volume
foreach run ( $runs )
3dmerge -1blur_fwhm 8.0 -doall -prefix pb03.$subj.r$run.blur \
pb02.$subj.r$run.volreg+orig
end
# ------------------------------------------------------# create 'full_mask' dataset (union mask)
foreach run ( $runs )
3dAutomask -dilate 1 -prefix rm.mask_r$run pb03.$subj.r$run.blur+orig
end
# get mean and compare it to 0 for taking 'union'
3dMean -datum short -prefix rm.mean rm.mask*.HEAD

98

3dcalc -a rm.mean+orig -expr 'ispositive(a-0)' -prefix full_mask.$subj
# ------------------------------------------------------# scale each voxel time series to have a mean of 100
# (subject to maximum value of 200)
foreach run ( $runs )
3dTstat -prefix rm.mean_r$run pb03.$subj.r$run.blur+orig
3dcalc -a pb03.$subj.r$run.blur+orig -b rm.mean_r$run+orig \
-c full_mask.$subj+orig
\
-expr 'c * min(200, a/b*100)'
\
-prefix pb04.$subj.r$run.scale
end
# ------------------------------------------------------# run the regression analysis
# create -stim_times files
make_stim_times.py -prefix stim_times -tr 3.0 -nruns 1 -nt 60
\
-files stimuli/stim.1D
foreach run ( $runs )
3dDeconvolve -input pb04.$subj.r$run.scale+orig.HEAD
\
-polort 2
\
-mask full_mask.$subj+orig
\
-num_stimts 7
\
-stim_times 1 stimuli/stim_times.01.1D 'GAM'
\
-stim_label 1 stim
\
-stim_file 2 dfile.rall.1D'[0]' -stim_base 2 -stim_label 2 roll \
-stim_file 3 dfile.rall.1D'[1]' -stim_base 3 -stim_label 3 pitch \
-stim_file 4 dfile.rall.1D'[2]' -stim_base 4 -stim_label 4 yaw \
-stim_file 5 dfile.rall.1D'[3]' -stim_base 5 -stim_label 5 dS \
-stim_file 6 dfile.rall.1D'[4]' -stim_base 6 -stim_label 6 dL \
-stim_file 7 dfile.rall.1D'[5]' -stim_base 7 -stim_label 7 dP \
-fout -tout -x1D X.xmat.1D -xjpeg X.jpg
\
-fitts fitts.$subj.r$run
\
-bucket stats.$subj.r$run
end
# if 3dDeconvolve fails, terminate the script
if ( $status != 0 ) then
echo '---------------------------------------'
echo '** 3dDeconvolve error, failing...'
echo ' (consider the file 3dDeconvolve.err)'
exit
endif
# create an all_runs dataset to match the fitts, errts, etc.

99

3dTcat -prefix all_runs.$subj pb04.$subj.r??.scale+orig.HEAD
# create ideal files for each stim type
1dcat X.xmat.1D'[9]' > ideal_stim.1D
# ------------------------------------------------------# remove temporary rm.* files
\rm -f rm.*
#return to parent folder
cd ../
#Copy and paste functional runs
into 'func.tlrc' folder
#Register functional runs into talairaching view
set subj = AM5023
cd ../{$subj}.results/
cp stats.{$subj}* ../func.tlrc/
cd ../../anatomy/
cp SPGR* ../Sec1/func.tlrc/
adwarp -apar SPGR+tlrc -dpar stats.{$subj}.r01+orig
adwarp -apar SPGR+tlrc -dpar stats.{$subj}.r02+orig
adwarp -apar SPGR+tlrc -dpar stats.{$subj}.r03+orig
#! now the data analysis for single subject's dataset is done. It is ready for group analysis!
#ANOVA
#this script analyzes pain intensity by subject as a random factor
#COMMENT: this script analyzes pain intensity by subject as a random factor
#MAKE SURE YOU HAVE blevels SET TO THE CORRECT NUMBER OF SUBJECTS AND
THAT YOU
#HAVE THE CORRECT RUN NAME AND PATH FOR EACH SUBJECT
#USE THE WARP AND BLURRED SUBJECT DATA AS INPUT
#YOU SHOULD HAVE 2XTHE NUMBER OF SUBJECTS LINES, EACH DESIGNATED
BY -dset
#sad session with 3 runs: pain only, sad picture only and sad+pain
3dANOVA2 -type 3 -alevels 3 -blevels 15 \
-dset 1 1 ../../AM5023/Sec1/func.tlrc/stats.AM5023.r01+tlrc"[1]" \
-dset 2 1 ../../AM5023/Sec1/func.tlrc/stats.AM5023.r02+tlrc"[1]" \
-dset 3 1 ../../AM5023/Sec1/func.tlrc/stats.AM5023.r03+tlrc"[1]" \
-dset 1 2 ../../Altered_Pain_Control_Subjects/EG5037/Sec1/func.tlrc/stats.EG5037.r01+tlrc"[1]"
\
-dset 2 2 ../../EG5037/Sec1/func.tlrc/stats.EG5037.r02+tlrc"[1]" \
-dset 3 2 ../../EG5037/Sec1/func.tlrc/stats.EG5037.r03+tlrc"[1]" \
-dset 1 3 ../../HG5044/Sec1/func.tlrc/stats.HG5044.r01+tlrc"[1]" \
-dset 2 3 ../../HG5044/Sec1/func.tlrc/stats.HG5044.r02+tlrc"[1]" \

100

-dset 3 3 ../../HG5044/Sec1/func.tlrc/stats.HG5044.r03+tlrc"[1]" \
-dset 1 4 ../../LM5057/Sec1/func.tlrc/stats.LM5057.r01+tlrc"[1]" \
-dset 2 4 ../../LM5057/Sec1/func.tlrc/stats.LM5057.r02+tlrc"[1]" \
-dset 3 4 ../../LM5057/Sec1/func.tlrc/stats.LM5057.r03+tlrc"[1]" \
-dset 1 5 ../../NM5022/Sec2/func.tlrc/stats.NM5022.r01+tlrc"[1]" \
-dset 2 5 ../../NM5022/Sec2/func.tlrc/stats.NM5022.r02+tlrc"[1]" \
-dset 3 5 ../../NM5022/Sec2/func.tlrc/stats.NM5022.r03+tlrc"[1]" \
-dset 1 6 ../../EP5047/Sec2/func.tlrc/stats.EP5047.r01+tlrc"[1]" \
-dset 2 6 ../../EP5047/Sec2/func.tlrc/stats.EP5047.r02+tlrc"[1]" \
-dset 3 6 ../../EP5047/Sec2/func.tlrc/stats.EP5047.r03+tlrc"[1]" \
-dset 1 7 ../../KM5039/Sec2/func.tlrc/stats.KM5039.r01+tlrc"[1]" \
-dset 2 7 ../../KM5039/Sec2/func.tlrc/stats.KM5039.r02+tlrc"[1]" \
-dset 3 7 ../../KM5039/Sec2/func.tlrc/stats.KM5039.r03+tlrc"[1]" \
-dset 1 8 ../../LY5043/Sec2/func.tlrc/stats.LY5043.r01+tlrc"[1]" \
-dset 2 8 ../../LY5043/Sec2/func.tlrc/stats.LY5043.r02+tlrc"[1]" \
-dset 3 8 ../../LY5043/Sec2/func.tlrc/stats.LY5043.r03+tlrc"[1]" \
-dset 1 9 ../../KM5053/Sec1/func.tlrc/stats.KM5053.r01+tlrc"[1]" \
-dset 2 9 ../../KM5053/Sec1/func.tlrc/stats.KM5053.r02+tlrc"[1]" \
-dset 3 9 ../../KM5053/Sec1/func.tlrc/stats.KM5053.r03+tlrc"[1]" \
-dset 1 10 ../../PN5045/Sec1/func.tlrc/stats.PN5045.r01+tlrc"[1]" \
-dset 2 10 ../../PN5045/Sec1/func.tlrc/stats.PN5045.r02+tlrc"[1]" \
-dset 3 10 ../../PN5045/Sec1/func.tlrc/stats.PN5045.r03+tlrc"[1]" \
-dset 1 11 ../../TW5058/Sec1/func.tlrc/stats.TW5058.r01+tlrc"[1]" \
-dset 2 11 ../../TW5058/Sec1/func.tlrc/stats.TW5058.r02+tlrc"[1]" \
-dset 3 11 ../../TW5058/Sec1/func.tlrc/stats.TW5058.r03+tlrc"[1]" \
-dset 1 12 ../../AK5046/Sec1/func.tlrc/stats.AK5046.r01+tlrc"[1]" \
-dset 2 12 ../../AK5046/Sec1/func.tlrc/stats.AK5046.r02+tlrc"[1]" \
-dset 3 12 ../../AK5046/Sec1/func.tlrc/stats.AK5046.r03+tlrc"[1]" \
-dset 1 13 ../../DP5031/Sec2/func.tlrc/stats.DP5031.r01+tlrc"[1]" \
-dset 2 13 ../../DP5031/Sec2/func.tlrc/stats.DP5031.r02+tlrc"[1]" \
-dset 3 13 ../../DP5031/Sec2/func.tlrc/stats.DP5031.r03+tlrc"[1]" \
-dset 1 14 ../../KJ5033/Sec2/func.tlrc/stats.KJ5033.r01+tlrc"[1]" \
-dset 2 14 ../../KJ5033/Sec2/func.tlrc/stats.KJ5033.r02+tlrc"[1]" \
-dset 3 14 ../../KJ5033/Sec2/func.tlrc/stats.KJ5033.r03+tlrc"[1]" \
-dset 1 15 ../../RT5026/Sec2/func.tlrc/stats.RT5026.r01+tlrc"[1]" \
-dset 2 15 ../../RT5026/Sec2/func.tlrc/stats.RT5026.r02+tlrc"[1]" \
-dset 3 15 ../../RT5026/Sec2/func.tlrc/stats.RT5026.r03+tlrc"[1]" \
#-dset 1 16 ../../TC5035/Sec2/func.tlrc/stats.TC5035.r01+tlrc"[1]" \
#-dset 2 16 ../../TC5035/Sec2/func.tlrc/stats.TC5035.r02+tlrc"[1]" \
#-dset 3 16 ../../TC5035/Sec2/func.tlrc/stats.TC5035.r03+tlrc"[1]" \
-fa stimuli \
-amean 1 neut \
-amean 2 sad_pics \
-amean 3 sad_pain \
-adiff 1 2 neut_v_pics \
-adiff 3 1 pain_v_neut \

101

-adiff 3 2 pain_v_pics \
-acontr -1 -1 1 sadpain_VS_neut_picsonly \
-acontr -1 -1 2 sadpain_v_neut_plus_sadpain_v_piconly \
-bucket anova.15Ss.avgNeut.MTLRC
#The steps for ROI analysis to make masks and get z-scores
#1. convert T score to z core from ANOVA results
3dmerge -doall -1zscore -prefix 15Ss.avgNeut.Zscore
anova.15Ss.avgNeut.MTLRC+tlrc'[3,5,7,9,11,13,15,17,19]'
3dbucket -prefix anova.15Ss.avgNeut.Zscore \
anova.15S.avgNeut.MTLRC+tlrc'[0]' anova.15S.avgNeut.MTLRC+tlrc'[1]' \
anova.15S.avgNeut.MTLRC+tlrc'[2]' 15Ss.avgNeut.Zscore+tlrc'[0]' \
anova.15S.avgNeut.MTLRC+tlrc'[4]' 15Ss.avgNeut.Zscore+tlrc'[1]' \
anova.15S.avgNeut.MTLRC+tlrc'[6]' 15Ss.avgNeut.Zscore+tlrc'[2]' \
anova.15S.avgNeut.MTLRC+tlrc'[8]' 15Ss.avgNeut.Zscore+tlrc'[3]' \
anova.15S.avgNeut.MTLRC+tlrc'[10]' 15Ss.avgNeut.Zscore+tlrc'[4]' \
anova.15S.avgNeut.MTLRC+tlrc'[12]' 15Ss.avgNeut.Zscore+tlrc'[5]' \
anova.15S.avgNeut.MTLRC+tlrc'[14]' 15Ss.avgNeut.Zscore+tlrc'[6]' \
anova.15S.avgNeut.MTLRC+tlrc'[16]' 15Ss.avgNeut.Zscore+tlrc'[7]'
#2. get clusters either from GUI of AFNI (save masks and cluster information as step 7) or from
the script as the below, and save masks
3dmerge -1thtoin -1noneg -1thresh 4.13 -1clust_order 1.7 128 -prefix
Mask8_9_4.13_.002_Vol200_rmm4 -1dindex 8 -1tindex 9
anova.alt.pain.11UsableSs.041107+tlrc
#3. for each mask, test the false positive of selected p value (optional)
AlphaSim \
-mask Mask8_9_3.58_Vol200_rmm4+tlrc \
-iter 1000 -rmm 4 -pthr .005 -fwhm 3 \
-out SadPicsOnly.Mask.MonteCarlo.005.12_13
#4. use -3dclust to get the ranges of x,y,z; manual select coordinates and make masks for each
cluster (RAI-DICOM order), use excel to sort data and get rid of unnecessary coordinates
3dmaskdump -mask NeutPain_1.0E4_mask+tlrc -o NeutPain_1.0E4 -noijk -xyz -nozero
anova.14Ss.avgNeut.Zscore+tlrc'[3]'
#5. assembles a 3D dataset from an ASCII list of coordinates
3dUndump -prefix LSII_NeutPain -master 14Ss.avgNeut.Zscore+tlrc -datum float -xyz LSII.txt
#6. find peaks of clusters
(optional)
3dExtrema -output SadPicsOlny_peaks -mask_file Mask8_9_3.58_Vol200_rmm4+tlrc -maxima
-interior -slice -average anova.alt.pain.11UsableSs.041107+tlrc'[8]'

102

#7. get the summary of cluster (mean, coordinate, peak value)
3dclust -1Dformat -1noneg -1thresh 2.97 1.7 128 anova.14Ss.avgNeut.Zscore+tlrc'[7]' >
SadPain_Zscore2.97_cluster_rmm1.5_p003.1D
#optional
3dROIstats -mask Mask2_3SPic_t4.51_.002_Vol170_rmm1.8+tlrc -minmax -sigma -nzmean
'anova.alt.Sadpain.9Ss.Zscore.181007+tlrc[3]' > Spics.ave.1D

103

APPENDIX B

Programming scripts in AFNI for processing the imaging data of Chapter 3
#To process the resting-state fMRI data
#prepare dataset of each subject for simple correlation and ICA group analysis
# register 2-D images to 3-D AFNI data format
#!/bin/csh -f
to3d -session Afni_analy -prefix RestEyeClose -epan -time:zt 36 164 2500 alt+z
'./RestEyeClose/*'
to3d -session Afni_analy -prefix T1Volume './T1Volume/MRDC.*'
#pre-processing dataset of each subject
3dTshift -prefix ts_RestEyeClose+orig RestEyeClose+orig
3dvolreg -dfile mot_tempRestEyeClose -base 'ts_RestEyeClose+orig[163]' -prefix
reg_RestEyeClose ts_RestEyeClose+orig
3dmerge -1blur_fwhm 4 -doall -prefix reg_RestEyeClose_blur reg_RestEyeClose+orig
3dcalc -prefix RestEyeClose_mask800 -a 'ts_RestEyeClose+orig[163]' -expr 'astep(a,800)'
3dDeconvolve -input reg_RestEyeClose_blur+orig -polort 2 -num_stimts 6 \
-stim_file 1 'mot_tempRestEyeClose[1]' -stim_label 1 roll \
-stim_file 2 'mot_tempRestEyeClose[2]' -stim_label 2 pitch \
-stim_file 3 'mot_tempRestEyeClose[3]' -stim_label 3 yaw \
-stim_file 4 'mot_tempRestEyeClose[4]' -stim_label 4 IS \
-stim_file 5 'mot_tempRestEyeClose[5]' -stim_label 5 RL \
-stim_file 6 'mot_tempRestEyeClose[6]' -stim_label 6 AP \
-mask RestEyeClose_mask800+orig -fout -tout \
-cbucket cbucket_coef_RestEyeClose -x1D x1Dmatrix_RestEyeClose \
-bucket deconv_reg_RestEyeClose
#make clean data for simple correlation analysis
3dSynthesize -cbucket cbucket_coef_RestEyeClose+orig -matrix x1Dmatrix_RestEyeClose select all -prefix EffectsOfNoInterest_RestEyeClose
3dcalc -a reg_RestEyeClose_blur+orig -b EffectsOfNoInterest_RestEyeClose+orig -expr 'a-b' prefix CleanData_RestEyeClose
adwarp -apar T1Volume+tlrc -dpar 'CleanData_RestEyeClose+orig' -dxyz 3 -prefix
CleanData_RestEyeClose_dxyz3_man
adwarp -apar T1Volume+tlrc -dpar 'RestEyeClose_mask800+orig' -dxyz 3 -prefix
RestEyeClose_mask800_dxyz3_man
3dmaskave -quiet -mask RestEyeClose_mask800_dxyz3_man+tlrc
CleanData_RestEyeClose_dxyz3_man+tlrc > CleanData_RestEyeClose_dxyz3_man_global.1D
##Simple Correlation analysis
#For each defined ROI of each subject, calculate correlation coefficients (use PPA_left as an
example)

104

adwarp -apar T1Volume+tlrc -dpar ROI_PPA_Left+orig -dxyz 3 -prefix
ROI_PPA_Left_dxyz3_man
3dcalc -a ROI_PPA_Left_dxyz3_man+tlrc -b RestEyeClose_mask800_dxyz3_man+tlrc -expr
'a*b' -prefix ROI_PPA_Left_dxyz3_man_mask800
3dmaskave -quiet -mask ROI_PPA_Left_dxyz3_man_mask800+tlrc
CleanData_RestEyeClose_dxyz3_man+tlrc >
Seed_CleanData_RestEyeClose_dxyz3_man_PPA_Left.1D
3dDeconvolve -input CleanData_RestEyeClose_dxyz3_man+tlrc -polort 0 -num_stimts 2 \
-stim_file 1 CleanData_RestEyeClose_dxyz3_man_global.1D -stim_label 1
RegressorOfNoInterest \
-stim_file 2 Seed_CleanData_RestEyeClose_dxyz3_man_PPA_Left.1D -stim_label 2
CorrCoef -tout -rout -fitts fit_RestEyeClose_PPA_Left -bucket
Corr_RestEyeClose_PPA_Left_3dDeconvolve

## For the group of 16 subjects
set subject = (RS_001 RS_002 RS_003 RS_005 RS_008 RS_009 RS_010 RS_014 RS_015
RS_020 RS_021 RS_022 RS_023 RS_027 RS_028 RS_026)
set ROIs = (SI_Left SII_Left ACC_Left PInsula_Left PInsula_Right MedialFro RInfFro
Hippo_Right Hippo_Left LPAG LBrainstem)
foreach sub($subject)
echo $sub
cd ../{$sub}/Afni_analy/Corr_Analysis/
#take square root of the output from 3dDeconvole to get R
#convert correlation coefficient R to Gaussian
foreach ROI($ROIs)
echo $ROI
gunzip Corr_RestEyeClose_{$ROI}_3dDeconvolve+tlrc.BRIK.gz
3dcalc -a Corr_RestEyeClose_{$ROI}_3dDeconvolve+tlrc'[7]' -b
Corr_RestEyeClose_{$ROI}_3dDeconvolve+tlrc'[5]' -expr 'ispositive(b)*sqrt(a)isnegative(b)*sqrt(a)' -prefix Corr_RestEyeClose_{$ROI}_R
3dcalc -a Corr_RestEyeClose_{$ROI}_R+tlrc -expr 'log((1+a)/(1-a))/2' -prefix
corr_{$sub}_{$ROI}_Z
mv corr_{$sub}_{$ROI}_Z* ../../../Corr_Group_Ana/
end
cd ../../../Corr_Group_Ana/
end
#get Z scores or regressor coefficients of all subjects
#covert z-score to t value only indicates r is with some significance level different from 0, does
not necessarily mean a high correlation

105

foreach ROI($ROIs)
3dttest -prefix Grp_{$ROI}_Result -base1 0 \
-set2 \
corr_RS_001_{$ROI}_Z+tlrc \
corr_RS_002_{$ROI}_Z+tlrc \
corr_RS_003_{$ROI}_Z+tlrc \
corr_RS_005_{$ROI}_Z+tlrc \
corr_RS_008_{$ROI}_Z+tlrc \
corr_RS_009_{$ROI}_Z+tlrc \
corr_RS_010_{$ROI}_Z+tlrc \
corr_RS_014_{$ROI}_Z+tlrc
end
foreach ROI($ROIs)
3dcalc -a Grp_{$ROI}_Result+tlrc'[0]' -expr '(exp(2*a)-1)/(exp(2*a)+1)' -prefix
Rvalue_Grp_{$ROI}
end
##ICA analysis
#Step 1 preprocess data, reduce functional data dimension to 25 using PCA
#use time-shift, motion corrected, smoothed and normalized data
set subject = (RS_001 RS_002 RS_003 RS_005 RS_008 RS_009 RS_010 RS_014 RS_015
RS_020 RS_021 RS_022 RS_023 RS_026 RS_027 RS_028)
foreach sub($subject)
echo $sub
cd ../{$sub}/Afni_analy/ICA/
#clean data reduce the correlation
3dDeconvolve -input reg_RestEyeClose_blur+orig -polort 2 -num_stimts 6 \
-stim_file 1 'mot_tempRestEyeClose[1]' -stim_label 1 roll \
-stim_file 2 'mot_tempRestEyeClose[2]' -stim_label 2 pitch \
-stim_file 3 'mot_tempRestEyeClose[3]' -stim_label 3 yaw \
-stim_file 4 'mot_tempRestEyeClose[4]' -stim_label 4 IS \
-stim_file 5 'mot_tempRestEyeClose[5]' -stim_label 5 RL \
-stim_file 6 'mot_tempRestEyeClose[6]' -stim_label 6 AP \
-mask RestEyeClose_mask800+orig -fout -tout \
-cbucket cbucket_coef_RestEyeClose -x1D x1Dmatrix_RestEyeClose \
-bucket deconv_reg_RestEyeClose
3dSynthesize -cbucket cbucket_coef_RestEyeClose+orig -matrix x1Dmatrix_RestEyeClose select all -prefix EffectsOfNoInterest_RestEyeClose
3dcalc -a reg_RestEyeClose_blur+orig -b EffectsOfNoInterest_RestEyeClose+orig -expr 'a-b' prefix CleanData_RestEyeClose

106

adwarp -apar T1Volume+tlrc -dpar 'CleanData_RestEyeClose+orig' -dxyz 3 -prefix CleanData
gunzip CleanData+tlrc.BRIK.gz
adwarp -apar T1Volume+tlrc -dpar 'RestEyeClose_mask800+orig' -dxyz 3 -prefix mask800
gunzip mask800+tlrc.BRIK.gz
# reduce time points for each subject from 164 to 25 using PCA
3dpc -vmean -vnorm -pcsave 25 -float -mask mask800+tlrc -prefix PreICA_TP25_{$sub}
CleanData+tlrc
mv PreICA_TP25_{$sub}* ../../../ICA_Group_Ana/
cd ../../../ICA_Group_Ana/
end
#step 2 concatenate data from all subjects
#copy mask800+tlrc* from RS_008 to current group analysis folder
3dTcat -session . -prefix Group_PreICA_concat_before_dimreduce PreICA_TP25_RS_001+tlrc
PreICA_TP25_RS_002+tlrc PreICA_TP25_RS_003+tlrc PreICA_TP25_RS_005+tlrc
PreICA_TP25_RS_008+tlrc PreICA_TP25_RS_009+tlrc PreICA_TP25_RS_010+tlrc
PreICA_TP25_RS_014+tlrc PreICA_TP25_RS_015+tlrc PreICA_TP25_RS_010+tlrc
PreICA_TP25_RS_021+tlrc PreICA_TP25_RS_022+tlrc PreICA_TP25_RS_023+tlrc
PreICA_TP25_RS_026+tlrc PreICA_TP25_RS_027+tlrc PreICA_TP25_RS_028+tlrc

##reduce dimension of aggregated data to 20 (number of sources selected)
3dpc -mask mask800+tlrc -reduce 20 Group_PreICA_concat_redim_RestEyeClose -prefix
PreICA_RestEyeClose Group_PreICA_concat_before_dimreduce+tlrc
3drename Group_PreICA_concat_redim_RestEyeClose ICAInput_RestEyeClose
# find the components
# .gz file can not be processed, needs to be unziped
gunzip ICAInput_RestEyeClose+tlrc.BRIK.gz
3dICA.R RestEyeClose_ICAOutput

##to run 3dICA.R, a par.txt file is needed to be saved in the same folder
## par.txt
Input:ICAInput_RestEyeClose+tlrc
CompOutput: RestEyeClose_ICAOutput
MixOutput:Temp
NoComp:20

107

Func:logcosh
Type:parallel

##Preprocess dataset for the path analysis
#step 1: define ROIs (saved in \remote\symonds\synapse3\FromeDavidZhu\mask)
# one portion of ROIs are drew according to anatomical definition manually
#in each subjects Corr_analysis folder, run 1deval to covert averaged time series of each ROI
into percent signal changes. (extROI.s)
#extract timeseries from masks. The masks are obtained from group analysis of altered pain
study (use the ROI of Left ACC as the example)
set subject = (RS_001 RS_002 RS_003 RS_005 RS_008 RS_009 RS_010 RS_014 RS_015
RS_020 RS_021 RS_022 RS_023 RS_026 RS_027 RS_028)
foreach sub($subject)
cd ../{$sub}/Afni_analy/Corr_Analysis/
1deval -expr 'step(a/b)*min(200,(a/b)*100)+step(-a/b)*max(-200,(a/b)*100)' -a
Seed_CleanData_RestEyeClose_dxyz3_man_ACC_Left.1D -b
CleanData_RestEyeClose_dxyz3_man_global.1D > {$sub}.per_clean_LACC24.1D
mv {$sub}.per_clean*.1D ../../../SEM
cd ../../../SEM/
end
#the other portions of ROIs were drew manually (extROI2.csh)
# Use the ROI of Left subgenual ACC(25) as the example
set subject = (RS_001 RS_002 RS_003 RS_005 RS_008 RS_009 RS_010 RS_014 RS_015
RS_020 RS_021 RS_022 RS_023 RS_026 RS_027 RS_028)
foreach sub($subject)
cd ../masks/
cp ROI*_dxyz3_man* ../{$sub}/Afni_analy/Corr_Analysis/
cd ../SEM/
end
foreach sub($subject)
cd ../{$sub}/Afni_analy/Corr_Analysis/

#Left subgenual ACC(25)
rm ROI_LACC25_dxyz3_man_mask800*
rm Seed_CleanData_RestEyeClose_dxyz3_man_LACC25.1D
3dcalc -a ROI_LACC25_dxyz3_man+tlrc -b RestEyeClose_mask800_dxyz3_man+tlrc -expr
'a*b' -prefix ROI_LACC25_dxyz3_man_mask800

108

3dmaskave -quiet -mask ROI_LACC25_dxyz3_man_mask800+tlrc
CleanData_RestEyeClose_dxyz3_man+tlrc >
Seed_CleanData_RestEyeClose_dxyz3_man_LACC25.1D
1deval -expr 'step(a/b)*min(200,(a/b)*100)+step(-a/b)*max(-200,(a/b)*100)' -a
Seed_CleanData_RestEyeClose_dxyz3_man_LACC25.1D -b
CleanData_RestEyeClose_dxyz3_man_global.1D > {$sub}.per_clean_LACC25.1D
mv {$sub}.per_clean*.1D ../../../SEM/
cd ../../../SEM/
end

# to perform path analysis, extract time series from ROIs
#extract timeseries from masks. The masks are the same as those used in resting state fMRI
#For pain network, ROIs are left left SI, left SII, right SII, R & L posterior Insula, L superior
temporal lobe, L ACC(BA24), L PAG, L VLPthalamus, L PCC and L hypothalamus (0 voxels
survives in hypothalamus, so get rid of this ROI)
#use the subjects saved in the folder of Control sec1 as the example (Neutral Pain runs (r01)
set subject = (AM5023 EG5037 HG5044 LM5057)
foreach sub($subject)
cp ../mask/ROI_LSI_mask* ../../../New_Altered_Pain/Altered_Pain_Control_Subjects/{$sub}/Se
c1/{$sub}.results/
cp ../mask/ROI_LPIn_mask* ../../../New_Altered_Pain/Altered_Pain_Control_Subjects/{$sub}/S
ec1/{$sub}.results/
cp ../mask/ROI_PAG_mask* ../../../New_Altered_Pain/Altered_Pain_Control_Subjects/{$sub}/S
ec1/{$sub}.results/
cp ../mask/ROI_LVPLTha_mask* ../../../New_Altered_Pain/Altered_Pain_Control_Subjects/{$s
ub}/Sec1/{$sub}.results/
cp ../mask/ROI_LPCC_mask* ../../../New_Altered_Pain/Altered_Pain_Control_Subjects/{$sub}/
Sec1/{$sub}.results/
cd ../../../New_Altered_Pain/Altered_Pain_Control_Subjects/{$sub}/Sec1/{$sub}.results/
rm roi*
rm *timeseries*

# register motion correced and blurd funcitonal run into tlrc
#get percent signal changes and extract time series
3dmaskave -quiet -mask full_mask.{$sub}+orig pb03.{$sub}.r01.blur+orig >
r01.blur.orig.global.1D
#right SII

109

3dfractionize -prefix roi_neut_0001_mask_{$sub} -template pb03.{$sub}.r01.blur+orig -warp
SPGR+tlrc -input Grp_15Ss_neut_0001_mask+tlrc -preserve -clip 0.5
3dmaskave -quiet -mask roi_neut_0001_mask_{$sub}+orig -mrange 1 1
pb03.{$sub}.r01.blur+orig > {$sub}_Npain1_RSII_timeseries.1D
1deval -expr 'step(a/b)*min(200,(a/b)*100)+step(-a/b)*max(-200,(a/b)*100)' -a
{$sub}_Npain1_RSII_timeseries.1D -b r01.blur.orig.global.1D > {$sub}_pain_RSII.1D
#right posterior insula
3dmaskave -quiet -mask roi_neut_0001_mask_{$sub}+orig -mrange 3 3
pb03.{$sub}.r01.blur+orig > {$sub}_Npain3_RPInsula_timeseries.1D
1deval -expr 'step(a/b)*min(200,(a/b)*100)+step(-a/b)*max(-200,(a/b)*100)' -a
{$sub}_Npain3_RPInsula_timeseries.1D -b r01.blur.orig.global.1D > {$sub}_pain_RPIn.1D
#left SII
3dfractionize -prefix roi_neut_001_mask_{$sub} -template pb03.{$sub}.r01.blur+orig -warp
SPGR+tlrc -input Grp_15Ss_neut_001_ACC_mask+tlrc -preserve -clip 0.5
3dmaskave -quiet -mask roi_neut_001_mask_{$sub}+orig -mrange 2 2
pb03.{$sub}.r01.blur+orig > {$sub}_Npain2_LSII_timeseries.1D
1deval -expr 'step(a/b)*min(200,(a/b)*100)+step(-a/b)*max(-200,(a/b)*100)' -a
{$sub}_Npain2_LSII_timeseries.1D -b r01.blur.orig.global.1D > {$sub}_pain_LSII.1D
#left ACC
3dmaskave -quiet -mask roi_neut_001_mask_{$sub}+orig -mrange 5 5
pb03.{$sub}.r01.blur+orig > {$sub}_Npain5_LACC_timeseries.1D
1deval -expr 'step(a/b)*min(200,(a/b)*100)+step(-a/b)*max(-200,(a/b)*100)' -a
{$sub}_Npain5_LACC_timeseries.1D -b r01.blur.orig.global.1D > {$sub}_pain_LACC.1D
#left superial tempral gyrus
3dmaskave -quiet -mask roi_neut_001_mask_{$sub}+orig -mrange 4 4
pb03.{$sub}.r01.blur+orig > {$sub}_Npain4_LSTC_timeseries.1D
1deval -expr 'step(a/b)*min(200,(a/b)*100)+step(-a/b)*max(-200,(a/b)*100)' -a
{$sub}_Npain4_LSTC_timeseries.1D -b r01.blur.orig.global.1D > {$sub}_pain_LSTG.1D
#left SI
3dfractionize -prefix roi_LSI_mask -template pb03.{$sub}.r01.blur+orig -warp SPGR+tlrc input ROI_LSI_mask+tlrc -clip 0.5
3dmaskave -quiet -mask roi_LSI_mask+orig pb03.{$sub}.r01.blur+orig >
{$sub}_LSI_timeseries.1D

110

1deval -expr 'step(a/b)*min(200,(a/b)*100)+step(-a/b)*max(-200,(a/b)*100)' -a
{$sub}_LSI_timeseries.1D -b r01.blur.orig.global.1D > {$sub}_pain_LSI.1D
#left posterior insula
3dfractionize -prefix roi_LPIn_mask -template pb03.{$sub}.r01.blur+orig -warp SPGR+tlrc input ROI_LPIn_mask+tlrc -clip 0.5
3dmaskave -quiet -mask roi_LPIn_mask+orig pb03.{$sub}.r01.blur+orig >
{$sub}_LPIn_timeseries.1D
1deval -expr 'step(a/b)*min(200,(a/b)*100)+step(-a/b)*max(-200,(a/b)*100)' -a
{$sub}_LPIn_timeseries.1D -b r01.blur.orig.global.1D > {$sub}_pain_LPIn.1D
#PAG
3dfractionize -prefix roi_PAG_mask -template pb03.{$sub}.r01.blur+orig -warp SPGR+tlrc input ROI_PAG_mask+tlrc -clip 0.5
3dmaskave -quiet -mask roi_PAG_mask+orig pb03.{$sub}.r01.blur+orig >
{$sub}_PAG_timeseries.1D
1deval -expr 'step(a/b)*min(200,(a/b)*100)+step(-a/b)*max(-200,(a/b)*100)' -a
{$sub}_PAG_timeseries.1D -b r01.blur.orig.global.1D > {$sub}_pain_PAG.1D
#left VPL thalamus
3dfractionize -prefix roi_LTha_mask -template pb03.{$sub}.r01.blur+orig -warp SPGR+tlrc input ROI_LVPLTha_mask+tlrc -clip 0.5
3dmaskave -quiet -mask roi_LTha_mask+orig pb03.{$sub}.r01.blur+orig >
{$sub}_LTha_timeseries.1D
1deval -expr 'step(a/b)*min(200,(a/b)*100)+step(-a/b)*max(-200,(a/b)*100)' -a
{$sub}_LTha_timeseries.1D -b r01.blur.orig.global.1D > {$sub}_pain_LTha.1D
#left posterior cingulate cortex
3dfractionize -prefix roi_LPCC_mask -template pb03.{$sub}.r01.blur+orig -warp SPGR+tlrc input ROI_LPCC_mask+tlrc -clip 0.5
3dmaskave -quiet -mask roi_LPCC_mask+orig pb03.{$sub}.r01.blur+orig >
{$sub}_LPCC_timeseries.1D
1deval -expr 'step(a/b)*min(200,(a/b)*100)+step(-a/b)*max(-200,(a/b)*100)' -a
{$sub}_LPCC_timeseries.1D -b r01.blur.orig.global.1D > {$sub}_pain_LPCC.1D

mv {$sub}_pain*.1D ../../../../SEM/Pain

111

cd ../../../../SEM/Pain
end

# combine all 1D files into one
set subject =(NM5022 EP5047 KM5039 LY5043 KM5053 PN5045 TW5058 AK5046 DP5031
KJ5033 RT5026)
foreach sub($subject)
1dMarry {$sub}_pain*.1D > {$sub}.pain.10ROIs.1D
end

# calculate mean of each ROIs across 15 Ss
set ROIs = (LACC LPCC LPIn LSI LSII LSTG LTha PAG RPIn RSII)
foreach roi($ROIs)
3dMean -prefix ts_mean_{$roi} *_pain_{$roi}.1D
end

#finally, put all the exacted time series into the software of AMOS in SPSS and perform the path
analysis.

112

REFERENCES

113

REFERENCES

Aggleton, J. P., M. J. Burton, et al. (1980). "Cortical and subcortical afferents to the amygdala of
the rhesus monkey (Macaca mulatta)." Brain Res 190(2): 347-68.
Albe-Fessard, D., K. J. Berkley, et al. (1985). "Diencephalic mechanisms of pain sensation."
Brain Res 356(3): 217-96.
Anand, A., Y. Li, et al. (2005). "Activity and connectivity of brain mood regulating circuit in
depression: a functional magnetic resonance study." Biol Psychiatry 57(10): 1079-88.
Apkarian, A. V., M. C. Bushnell, et al. (2005). "Human brain mechanisms of pain perception and
regulation in health and disease." Eur J Pain 9(4): 463-84.
Augustine, J. R. (1996). "Circuitry and functional aspects of the insular lobe in primates
including humans." Brain Res Brain Res Rev 22(3): 229-44.
Baliki, M. N., P. Y. Geha, et al. (2008). "Beyond feeling: chronic pain hurts the brain, disrupting
the default-mode network dynamics." J Neurosci 28(6): 1398-403.
Bär, K.-J., G. Wagner, et al. (2007). "Increased Prefrontal Activation During Pain Perception in
Major Depression." Biological Psychiatry 62(11): 1281-1287.
Barbas, H., S. Saha, et al. (2003). "Serial pathways from primate prefrontal cortex to autonomic
areas may influence emotional expression." BMC Neuroscience 4(1): 25.
Bard, P. and D. M. Rioch (1937). "A study of four cats deprived of neocortex and additional
portions of the forebrain." Bulletin of the Johns Hopkins Hospital 60: 73-153.
Beauregard, M., J. M. Leroux, et al. (1998). "The functional neuroanatomy of major depression:
an fMRI study using an emotional activation paradigm." Neuroreport 9(14): 3253-8.
Beck, A. T., R. A. Steer, et al. (1996). "Beck depression inventory." PsychCorp.
Bernard, J. F., H. Bester, et al. (1996). Chapter 14 Involvement of the spino-parabrachio amygdaloid and -hypothalamic pathways in the autonomic and affective emotional aspects of
pain. Progress in Brain Research, Elsevier. Volume 107: 243-255.
Blair, R. J., J. S. Morris, et al. (1999). "Dissociable neural responses to facial expressions of
sadness and anger." Brain 122 ( Pt 5): 883-93.
Blanchard, E. B., F. Andrasik, et al. (1985). "Behavioral treatment of 250 chronic headache
patients: A clinical replication series." Behavior Therapy 16(3): 308-327.

114

Breslau, N., R. B. Lipton, et al. (2003). "Comorbidity of migraine and depression: investigating
potential etiology and prognosis." Neurology 60(8): 1308-12.
Bushnell, M. C., G. H. Duncan, et al. (1999). "Pain perception: is there a role for primary
somatosensory cortex?" Proc Natl Acad Sci U S A 96(14): 7705-9.
Calejesan, A. A., S. J. Kim, et al. (2000). "Descending facilitatory modulation of a behavioral
nociceptive response by stimulation in the adult rat anterior cingulate cortex." Eur J Pain 4(1):
83-96.
Calhoun, V. D., T. Adali, et al. (2001). "A method for making group inferences from functional
MRI data using independent component analysis." Human Brain Mapping 14(3): 140-151.
Calhoun, V. D., T. AdalI, et al. (2004). "A method for comparing group fMRI data using
independent component analysis: application to visual, motor and visuomotor tasks." Magnetic
Resonance Imaging 22(9): 1181-1191.
Cannon, W. B. (1931). "Again the James-Lange and the thalamic theories of emotion."
Psychological Review 38: 281-295.
Cauda, F., K. Sacco, et al. (2009). "Altered resting state in diabetic neuropathic pain." PLoS One
4(2): e4542.
Cavada, C. and P. S. Goldman-Rakic (1989). "Posterior parietal cortex in rhesus monkey: II.
Evidence for segregated corticocortical networks linking sensory and limbic areas with the
frontal lobe." J Comp Neurol 287(4): 422-45.
Cheng, Y., C. P. Lin, et al. (2007). "Expertise Modulates the Perception of Pain in Others." Curr
Biol.
Coghill, R. C., D. J. Mayer, et al. (1993). "Wide dynamic range but not nociceptive-specific
neurons encode multidimensional features of prolonged repetitive heat pain." J Neurophysiol
69(3): 703-16.
Coghill, R. C., C. N. Sang, et al. (1999). "Pain intensity processing within the human brain: a
bilateral, distributed mechanism." J Neurophysiol 82(4): 1934-43.
Conti, F., M. Fabri, et al. (1988). "Immunocytochemical evidence for glutamatergic corticocortical connections in monkeys." Brain Res 462(1): 148-53.
Cox, R. W. (1996). "AFNI: software for analysis and visualization of functional magnetic
resonance neuroimages." Comput Biomed Res 29(3): 162-73.
Craig, A. D. (1995). "Distribution of brainstem projections from spinal lamina I neurons in the
cat and the monkey." J Comp Neurol 361(2): 225-48.

115

Craig, A. D. (2002). "How do you feel? Interoception: the sense of the physiological condition of
the body." Nat Rev Neurosci 3(8): 655-666.
Craig, A. D. (2003). "A new view of pain as a homeostatic emotion." Trends Neurosci 26(6):
303-7.
Craig, A. D. (2009). "How do you feel [mdash] now? The anterior insula and human awareness."
Nat Rev Neurosci 10(1): 59-70.
Craig, A. D., K. Chen, et al. (2000). "Thermosensory activation of insular cortex." Nat Neurosci
3(2): 184-90.
Critchley, H. D., S. Wiens, et al. (2004). "Neural systems supporting interoceptive awareness."
Nat Neurosci 7(2): 189-95.
Crown, E. D., T. E. King, et al. (2000). "Shock-induced hyperalgesia: III. Role of the bed
nucleus of the stria terminalis and amygdaloid nuclei." Behav Neurosci 114(3): 561-73.
Dagenais, S., J. Caro, et al. (2008). "A systematic review of low back pain cost of illness studies
in the United States and internationally." The Spine Journal 8(1): 8-20.
Dalgleish, T. (2004). "The emotional brain." Nat Rev Neurosci 5(7): 583-9.
Damasio, A. R., T. J. Grabowski, et al. (2000). "Subcortical and cortical brain activity during the
feeling of self-generated emotions." Nat Neurosci 3(10): 1049-56.
deCharms, R. C., F. Maeda, et al. (2005). "Control over brain activation and pain learned by
using real-time functional MRI." Proceedings of the National Academy of Sciences 102(51):
18626-18631.
DeLeo, J. A. (2006). "Basic Science of Pain." J Bone Joint Surg Am 88(suppl_2): 58-62.
Dennis, S. G. and R. Melzack (1977). "Pain-signalling systems in the dorsal and ventral spinal
cord." Pain 4(2): 97-132.
Derogatis, L. R. (1993). "Brief Symptom Inventory (BSI®)." PsychCorp.
Devinsky, O., M. J. Morrell, et al. (1995). "Contributions of anterior cingulate cortex to
behaviour." Brain 118 ( Pt 1): 279-306.
Dhond, R. P., C. Yeh, et al. (2008). "Acupuncture modulates resting state connectivity in default
and sensorimotor brain networks." Pain 136(3): 407-18.
Dong, W. K., L. D. Salonen, et al. (1989). "Nociceptive responses of trigeminal neurons in SII7b cortex of awake monkeys." Brain Res 484(1-2): 314-24.

116

Drevets, W. C. (2001). "Neuroimaging and neuropathological studies of depression: implications
for the cognitive-emotional features of mood disorders." Curr Opin Neurobiol 11(2): 240-9.
Fields, H. L. (2000). "Pain modulation: expectation, opioid analgesia and virtual pain." Prog
Brain Res 122: 245-53.
Finnegan, T. F., S. R. Chen, et al. (2005). "Effect of the {mu} opioid on excitatory and inhibitory
synaptic inputs to periaqueductal gray-projecting neurons in the amygdala." J Pharmacol Exp
Ther 312(2): 441-8.
Fox, M. D. and M. E. Raichle (2007). "Spontaneous fluctuations in brain activity observed with
functional magnetic resonance imaging." Nat Rev Neurosci 8(9): 700-11.
Fox, M. D., A. Z. Snyder, et al. (2005). "The human brain is intrinsically organized into dynamic,
anticorrelated functional networks." Proceedings of the National Academy of Sciences of the
United States of America 102(27): 9673-9678.
Frot, M., M. Magnin, et al. (2007). "Human SII and posterior insula differently encode thermal
laser stimuli." Cereb Cortex 17(3): 610-20.
Gallagher, R. M. (1999). "Treatment planning in pain medicine. Integrating medical, physical,
and behavioral therapies." Med Clin North Am 83(3): 823-49, viii.
Greenspan, J. D., R. R. Lee, et al. (1999). "Pain sensitivity alterations as a function of lesion
location in the parasylvian cortex." Pain 81(3): 273-282.
Greicius, M. D., B. Krasnow, et al. (2003). "Functional connectivity in the resting brain: a
network analysis of the default mode hypothesis." Proc Natl Acad Sci U S A 100(1): 253-8.
Gureje, O., G. E. Simon, et al. (2001). "A cross-national study of the course of persistent pain in
primary care." Pain 92(1-2): 195-200.
Hamann, S. B., T. D. Ely, et al. (1999). "Amygdala activity related to enhanced memory for
pleasant and aversive stimuli." Nat Neurosci 2(3): 289-93.
Hirshberg, R. M., E. D. Al-Chaer, et al. (1996). "Is there a pathway in the posterior funiculus that
signals visceral pain?" Pain 67(2-3): 291-305.
Hofbauer, R. K., P. Rainville, et al. (2001). "Cortical Representation of the Sensory Dimension
of Pain." J Neurophysiol 86(1): 402-411.
Hsieh, J. C., S. Stone-Elander, et al. (1999). "Anticipatory coping of pain expressed in the human
anterior cingulate cortex: a positron emission tomography study." Neurosci Lett 262(1): 61-4.
Jackson, P. L., E. Brunet, et al. (2006). "Empathy examined through the neural mechanisms
involved in imagining how I feel versus how you feel pain." Neuropsychologia 44(5): 752-61.

117

Johansen-Berg, H., D. A. Gutman, et al. (2008). "Anatomical connectivity of the subgenual
cingulate region targeted with deep brain stimulation for treatment-resistant depression." Cereb
Cortex 18(6): 1374-83.
Kandel, E., J. Schwartz, et al. (2000). "The perception of pain." Principles of Neural Science.
Katims, J. J. (1997). "Neuroselective current perception threshold quantitative sensory test."
Muscle Nerve 20(11): 1468-9.
Kenshalo, D. R., Jr. and O. Isensee (1983). "Responses of primate SI cortical neurons to noxious
stimuli." J Neurophysiol 50(6): 1479-96.
Kim, D.-H., E. Adalsteinsson, et al. (2002). "Regularized higher-order in vivo shimming."
Magnetic Resonance in Medicine 48(4): 715-722.
Klüver, H. and P. C. Bucy (1937). "Psychic blindness and other symptoms following bilateral
temporal lobectomy in rhesus monkeys." American Journal of Physiology 119: 352-353.
Lane, R. D., P. M. Chua, et al. (1999). "Common effects of emotional valence, arousal and
attention on neural activation during visual processing of pictures." Neuropsychologia 37(9):
989-97.
Lane, R. D., G. R. Fink, et al. (1997). "Neural activation during selective attention to subjective
emotional responses." Neuroreport 8(18): 3969-72.
Lang, P. J., M. M. Bradley, et al. (1998). "Emotional arousal and activation of the visual cortex:
an fMRI analysis." Psychophysiology 35(2): 199-210.
Lang, P. J., M. K. Greenwald, et al. (1993). "Looking at pictures: affective, facial, visceral, and
behavioral reactions." Psychophysiology 30(3): 261-73.
Larsen, R. J. (1984). "Theory and measurement of affect intensity as an individual difference
characteristics." Dissertation Abastracts International 85(2297B (University Microfilms No. 8422112)).
Leonard, C. M., E. T. Rolls, et al. (1985). "Neurons in the amygdala of the monkey with
responses selective for faces." Behavioural Brain Research 15(2): 159-176.
Liotti, M., H. S. Mayberg, et al. (2000). "Differential limbic-cortical correlates of sadness and
anxiety in healthy subjects: implications for affective disorders." Biological Psychiatry 48(1):
30-42.
Logothetis, N. K., J. Pauls, et al. (2001). "Neurophysiological investigation of the basis of the
fMRI signal." Nature 412(6843): 150-7.

118

MacLean, P. (1949). "Psychosomatic disease and the visceral brain; recent developments bearing
on the Papez theory of emotion." Psychosom Med 11(6): 338-53.
Maihofner, C., B. Herzner, et al. (2006). "Secondary somatosensory cortex is important for the
sensory-discriminative dimension of pain: a functional MRI study." Eur J Neurosci 23(5): 137783.
Manzoni, T., F. Conti, et al. (1986). "Callosal projections from area SII to SI in monkeys:
anatomical organization and comparison with association projections." J Comp Neurol 252(2):
245-63.
Mayberg, H. S., S. K. Brannan, et al. (2000). "Regional metabolic effects of fluoxetine in major
depression: serial changes and relationship to clinical response." Biol Psychiatry 48(8): 830-43.
Mayberg, H. S., P. J. Lewis, et al. (1994). "Paralimbic hypoperfusion in unipolar depression." J
Nucl Med 35(6): 929-34.
Mayberg, H. S., M. Liotti, et al. (1999). "Reciprocal limbic-cortical function and negative mood:
converging PET findings in depression and normal sadness." Am J Psychiatry 156(5): 675-82.
Mayer, D. J., D. D. Price, et al. (1975). "Neurophysiological characterization of the anterolateral
spinal cord neurons contributing to pain perception in man." Pain 1(1): 51-8.
McIntosh, A. R., Gonzalez-Lima, F. (1992). "The application of structural modeling to metabolic
mapping of functional neural systems." Kluwer Academic Publishers: 219-255.
McKeown, M. J., S. Makeig, et al. (1998). "Analysis of fMRI data by blind separation into
independent spatial components." Human Brain Mapping 6(3): 160-188.
Meagher, M. W., R. C. Arnau, et al. (2001). "Pain and emotion: effects of affective picture
modulation." Psychosom Med 63(1): 79-90.
Melzack, R. (1989). "Labat lecture. Phantom limbs." Reg Anesth 14(5): 208-11.
Melzack, R. (1990). "Phantom limbs and the concept of a neuromatrix." Trends Neurosci 13(3):
88-92.
Melzack, R. and K. Casey (1968). "Sensory, motivational and central control determinants of
chronic pain: A new conceptual model." In Kenshalo, DR. The Skin Senses. Springfield, Illinois:
Thomas: 432.
Melzack, R. and P. D. Wall (1965). "Pain Mechanisms: A New Theory." Science 150(3699):
971-979.
Merskey, H. and N. Bogduk (1994). "Classification of chronic pain." Seattle (WA): IASP Press.

119

Mesulam, M. M. and E. J. Mufson (1982). "Insula of the old world monkey. III: Efferent cortical
output and comments on function." J Comp Neurol 212(1): 38-52.
Miller, E. K. and J. D. Cohen (2001). "An integrative theory of prefrontal cortex function." Annu
Rev Neurosci 24: 167-202.
Mobascher, A., J. Brinkmeyer, et al. (2009). "Laser-evoked potential P2 single-trial amplitudes
covary with the fMRI BOLD response in the medial pain system and interconnected subcortical
structures." Neuroimage 45(3): 917-26.
Mollet, G. A. and D. W. Harrison (2006). "Emotion and pain: a functional cerebral systems
integration." Neuropsychol Rev 16(3): 99-121.
Morris, J. S., C. D. Frith, et al. (1996). "A differential neural response in the human amygdala to
fearful and happy facial expressions." Nature 383(6603): 812-5.
Mufson, E. J., M. M. Mesulam, et al. (1981). "Insular interconnections with the amygdala in the
rhesus monkey." Neuroscience 6(7): 1231-48.
Neal, J. W., R. C. Pearson, et al. (1987). "The cortico-cortical connections of area 7b, PF, in the
parietal lobe of the monkey." Brain Res 419(1-2): 341-6.
Neugebauer, V., W. Li, et al. (2004). "The amygdala and persistent pain." Neuroscientist 10(3):
221-34.
Ogawa, S., T. M. Lee, et al. (1990). "Brain magnetic resonance imaging with contrast dependent
on blood oxygenation." Proc Natl Acad Sci U S A 87(24): 9868-72.
Ogino, Y., H. Nemoto, et al. (2007). "Inner experience of pain: imagination of pain while
viewing images showing painful events forms subjective pain representation in human brain."
Cereb Cortex 17(5): 1139-46.
Oldfield, R. C. (1971). "The assessment and analysis of handedness: the Edinburgh inventory."
Neuropsychologia 9(1): 97-113.
Pandya, D. N., G. W. Van Hoesen, et al. (1981). "Efferent connections of the cingulate gyrus in
the rhesus monkey." Exp Brain Res 42(3-4): 319-30.
Papez, J. W. (1995). "A proposed mechanism of emotion. 1937." J Neuropsychiatry Clin
Neurosci 7(1): 103-12.
Pessoa, L., S. Padmala, et al. (2005). "Fate of unattended fearful faces in the amygdala is
determined by both attentional resources and cognitive modulation." NeuroImage 28(1): 249-255.
Peyron, R., L. Garcia-Larrea, et al. (1999). "Haemodynamic brain responses to acute pain in
humans: sensory and attentional networks." Brain 122 ( Pt 9): 1765-80.

120

Peyron, R., B. Laurent, et al. (2000). "Functional imaging of brain responses to pain. A review
and meta-analysis (2000)." Neurophysiol Clin 30(5): 263-88.
Phan, K. L., T. Wager, et al. (2002). "Functional neuroanatomy of emotion: a meta-analysis of
emotion activation studies in PET and fMRI." Neuroimage 16(2): 331-48.
Phan, K. L., T. D. Wager, et al. (2004). "Functional neuroimaging studies of human emotions."
CNS Spectr 9(4): 258-66.
Phillips, M. L., A. W. Young, et al. (1997). "A specific neural substrate for perceiving facial
expressions of disgust." Nature 389(6650): 495-498.
Porreca, F., M. H. Ossipov, et al. (2002). "Chronic pain and medullary descending facilitation."
Trends Neurosci 25(6): 319-25.
Price, D. D. (1999). "Psychological Mechanisms of Pain and Analgesia." Progress in Pain
Research and Management 15.
Price, D. D. (2000). "Psychological and neural mechanisms of the affective dimension of pain."
Science 288(5472): 1769-72.
Price, D. D., S. Long, et al. (1992). "Sensory testing of pathophysiological mechanisms of pain
in patients with reflex sympathetic dystrophy." Pain 49(2): 163-73.
Raichle, M. E., A. M. MacLeod, et al. (2001). "A default mode of brain function." Proc Natl
Acad Sci U S A 98(2): 676-82.
Rainville, P. (2002). "Brain mechanisms of pain affect and pain modulation." Curr Opin
Neurobiol 12(2): 195-204.
Rainville, P., G. H. Duncan, et al. (1997). "Pain affect encoded in human anterior cingulate but
not somatosensory cortex." Science 277(5328): 968-71.
Rauch, S. L., M. A. Jenike, et al. (1994). "Regional cerebral blood flow measured during
symptom provocation in obsessive-compulsive disorder using oxygen 15-labeled carbon dioxide
and positron emission tomography." Arch Gen Psychiatry 51(1): 62-70.
Reiman, E. M. (1997). "The application of positron emission tomography to the study of normal
and pathologic emotions." J Clin Psychiatry 58 Suppl 16: 4-12.
Reiman, E. M., R. D. Lane, et al. (1997). "Neuroanatomical correlates of externally and
internally generated human emotion." Am J Psychiatry 154(7): 918-25.
Ressler, K. J. and H. S. Mayberg (2007). "Targeting abnormal neural circuits in mood and
anxiety disorders: from the laboratory to the clinic." Nat Neurosci 10(9): 1116-24.

121

Reynolds, D. V. (1969). "Surgery in the Rat during Electrical Analgesia Induced by Focal Brain
Stimulation." Science 164(3878): 444-445.
Robinson, C. J. and H. Burton (1980). "Somatic submodality distribution within the second
somatosensory (SII), 7b, retroinsular, postauditory, and granular insular cortical areas of M.
fascicularis." The Journal of Comparative Neurology 192(1): 93-108.
Schweinhardt, P., N. Kalk, et al. (2008). "Investigation into the neural correlates of emotional
augmentation of clinical pain." Neuroimage 40(2): 759-66.
Semple, W. E., P. F. Goyer, et al. (2000). "Higher brain blood flow at amygdala and lower
frontal cortex blood flow in PTSD patients with comorbid cocaine and alcohol abuse compared
with normals." Psychiatry 63(1): 65-74.
Sewards, T. V. and M. A. Sewards (2002). "The medial pain system: Neural representations of
the motivational aspect of pain." Brain Research Bulletin 59(3): 163-180.
Singer, T., B. Seymour, et al. (2004). "Empathy for pain involves the affective but not sensory
components of pain." Science 303(5661): 1157-62.
Stahl, S. M. (2002). "Does depression hurt?" J Clin Psychiatry 63(4): 273-4.
Sudheimer, K., K. Davis, et al. (2001). "An emotion induction paradigm for neuroimaging:
International Affective Picture System (IAPS)." Society of Neuroscience Abstract 27.
Symonds, L. L., N. S. Gordon, et al. (2006). "Right-Lateralized Pain Processing in the Human
Cortex: An fMRI Study." J Neurophysiol 95(6): 3823-3830.
Tang, N. K., P. M. Salkovskis, et al. (2008). "Effects of mood on pain responses and pain
tolerance: An experimental study in chronic back pain patients." Pain.
Taylor, S. F., I. Liberzon, et al. (1998). "The effect of emotional content on visual recognition
memory: a PET activation study." Neuroimage 8(2): 188-97.
Tolle, T. R., T. Kaufmann, et al. (1999). "Region-specific encoding of sensory and affective
components of pain in the human brain: a positron emission tomography correlation analysis."
Ann Neurol 45(1): 40-7.
Treede, R. D., D. R. Kenshalo, et al. (1999). "The cortical representation of pain." Pain 79(2-3):
105-11.
Valet, M., T. Sprenger, et al. (2004). "Distraction modulates connectivity of the cingulo-frontal
cortex and the midbrain during pain--an fMRI analysis." Pain 109(3): 399-408.

122

Villemure, C., S. Wassimi, et al. (2006). "Unpleasant odors increase pain processing in a patient
with neuropathic pain: psychophysical and fMRI investigation." Pain 120(1-2): 213-20.
Vogt, B. A. (2005). "Pain and emotion interaction in subregions of the cingulate gyrus." Nat Rev
Neurosci 6(7): 533-44.
Vogt, B. A. and R. W. Sikes (2000). "The medial pain system, cingulate cortex, and parallel
processing of nociceptive information." Prog Brain Res 122: 223-35.
Vogt, B. A., R. W. Sikes, et al. (1999). Chapter 16 The medial pain system, cingulate cortex, and
parallel processing of nociceptive information. Progress in Brain Research, Elsevier. Volume
122: 223-235.
Vuilleumier, P., M. P. Richardson, et al. (2004). "Distant influences of amygdala lesion on visual
cortical activation during emotional face processing." Nat Neurosci 7(11): 1271-8.
Weisenberg, M., T. Raz, et al. (1998). "The influence of film-induced mood on pain perception."
Pain 76(3): 365-75.
Weiskrantz, L. (1956). "Behavioral changes associated with ablation of the amygdaloid complex
in monkeys." J Comp Physiol Psychol 49(4): 381-91.
Wiech, K., M. Farias, et al. (2008). "An fMRI study measuring analgesia enhanced by religion as
a belief system." Pain 139(2): 467-476.
Willis, W. D. (1985). "Nociceptive pathways: anatomy and physiology of nociceptive ascending
pathways." Philos Trans R Soc Lond B Biol Sci 308(1136): 253-70.
Willis, W. D. and K. N. Westlund (1997). "Neuroanatomy of the Pain System and of the
Pathways That Modulate Pain." Journal of Clinical Neurophysiology 14(1): 2-31.
Willoughby, S. G., B. J. Hailey, et al. (2002). "The effect of laboratory-induced depressed mood
state on responses to pain." Behav Med 28(1): 23-31.
Yoshino, A., Y. Okamoto, et al. (2010). "Sadness enhances the experience of pain via neural
activation in the anterior cingulate cortex and amygdala: an fMRI study." Neuroimage 50(3):
1194-201.
Young, A. W., J. P. Aggleton, et al. (1995). "Face processing impairments after amygdalotomy."
Brain 118 ( Pt 1): 15-24.
Zhang, E. T. and A. D. Craig (1997). "Morphology and distribution of spinothalamic lamina I
neurons in the monkey." J Neurosci 17(9): 3274-84.
Zubieta, J. K., Y. R. Smith, et al. (2001). "Regional mu opioid receptor regulation of sensory and
affective dimensions of pain." Science 293(5528): 311-5.

123

Zubieta, J.-K., T. A. Ketter, et al. (2003). "Regulation of Human Affective Responses by
Anterior Cingulate and Limbic {micro}-Opioid Neurotransmission." Arch Gen Psychiatry
60(11): 1145-1153.

124