Neuronal processing of food pictures
Wissenschaftliche Arbeit
zur Erlangung des Grades eines Diplom‐Psychologen im Fachbereich Psychologie
der Universität Konstanz
vorgelegt von
Christoph Alexander Becker Schützenstr. 18
78462 Konstanz
Erstgutachter: Professor Dr. Harald Schupp Zweitgutachterin: Professorin Dr. Britta Renner
09 Konstanz, im August 20
Acknowledgements
I would like to thank the following persons: Prof. Dr. Harald Schupp and Prof. Dr.
Britta Renner for their willingness to supervise this work and their helpful guidance, Dr.
Tobias Flaisch and Frank Häcker for assistance during data acquisition and useful support, Dr. Ralf Schmälzle, Katrin Höhne and Anne Rösser for proofreading, all my friends for their social support and occasional scattering, and especially my parents, who put me through college and finally allowed me to perform this thesis.
Table of contents
Abstract ... 4
Introduction ... 5
Face and Object processing at the inferior temporal cortex ... 7
The pathway from sensation to cognition ... 5
Brain regions involved in visual food perception ... 9
A short overview on research with visual food cues ... 10
Building up a distributed network for food processing ... 14
Motivation and salience of picture categories ... 15
Influences on food perception ... 15
Deduction of thesis ... 16
Methods ... 17
Stimulus Material ... 17
Subjects ... 17
Data acquisition ... 18
Ex Da perimental Procedure ... 18
ta preprocessing and normalization ... 19
Preprocessing of functional data ... 20
Normalization of structural data ... 20
Normalization of functional data ... 20
Results ... 21
Re Results of questionnaires ... 21
sults of functional data ... 22
Effects of picture category ... 23
Contrasting picture categories against their scrambled counterparts ... 27
The influence of color ... 31
Discussion ... 39
Further effects of picture category ... 41
Effects of picture category at the inferior temporal cortex ... 40
Large scale distributed food network ... 43
Picture categories against their scrambled counterparts ... 43
The influence of color ... 45
Limitations of the present study ... 48
Future directions ... 48
Bibliography ... 50
Appendix ... 54
Sti Abbreviations ... 54
Usmulus material ... 55
ed questionnaires ... 58
Personal data ... 58
Hunger and appetite ... 59
Feelings with the scanner environment ... 60
Abstract
The aim of this study was to explore the neuronal representations that underlie the perception and recognition of visual presented food cues in black and white as well as colored form when contrasted against face, house, and their scrambled counterparts. A theoretical framework published by Mesulam (1998) was used that stated that neuronal concepts and higher order cognitive functions are represented in large scale distributed networks across the human cortex. These functions are provided by transmodal areas, which bind sensory specific properties of object classes into coherent conscious representation. The sensory representations in turn are stored in modality specific areas that represent this content with a gradient from low‐level feature encoding up to abstract object encoding. Through this framework a distributed food network was defined that should be responsible for the encoding of visual food cues, the anticipation of food’s tastiness, and the reward evaluation of food intake.
The results revealed that faces activate the fusiform face area (FFA), prefrontal regions, and parieto‐occipital cortex. On the other hand responded the parahippocampal place area (PPA) to houses. Food pictures activated the prefrontal cortex, orbitofrontal cortex, insula, inferior temporal cortex, and parieto‐occipital cortex. Moreover, object categories contrasted against their scrambled counterparts confirmed the results found through the contrasts reported above. An influence of color was found in prefrontal cortex, motor cortex, cingulate cortex, parieto‐occipital cortex, insula, and subcortical regions across the different contrasts.
The results of this study support the assumption that visual food cues are represented in a large scale distributed network across the human cortex. In this network the inferior temporal cortex is responsible for food object encoding, the insula for the processing of food’s anticipated tastiness, the orbitofrontal cortex for expected reward evaluation, prefrontal regions for action planning, the parieto‐occipital cortex for spatial attention processes, and the cingulate cortex for emotional salience. Further are the activation of FFA through faces and that of the PPA through houses consistent with findings in the literature. The influence of color showed that activity within regions responsible for action planning, action execution, motivation, attention, and emotional
wareness is altered.
a
Introduction
Food plays a large role in humans every day life and therefore it is understandable that a huge amount of research has focused on this topic. The human’s identification of food is in most cases first achieved through the visual modality by which an analysis of quality and acceptability of and based on this a decision for special foods is performed (Blackwell, 1995).
Ongoing on this assumption a part of food research has focused on visual presented food cues. In the neuro‐scientific domain the study of underlying brain structures, which are involved in the perception and processing of food, was aimed. Several studies have centered on this purpose in conjunction with differing topics, e.g. food’s calorie content (Killgore, et al., 2003), hunger/satiety (Hinton, et al., 2004), affect (Killgore & Yurgelun‐
Todd, 2006), and obesity (Martin, et al., 2009).
Whereas these topics take a large part in the literature, just a minority quested on the neuronal representation of food by itself, for example see a 2005 study (St‐Onge, Sy, Heymsfield, & Hirsch). However, even basic research is necessary to understand the principles of food related brain activations especially without confounding factors like calorie content, unge /satiety, affect, or obesity. h r
This study aims to identify the underlying neuronal representation of visual presented food without confounding of different factors as reported above. Therefore, a general framework of the pathway from sensation to cognition will be presented. This framework will then be linked to fundamental findings in visual object processing and thereafter to current findings of visual food research. After a short overview of former food research, a distributed network for food processing will be set up. Finally, salience of and effects on picture categories are introduced. Based on this some thesis will be deviated.
The pathway from sensation to cognition
As stated by Mesulam (1998) the pathway from sensation of a stimulus to its representation in memory and mind can be subdivided into hierarchical, distributed, and parallel and serial processing areas. The hierarchical organization can be characterized by the areas degree of synaptic levels. Higher order creatures have a more complex central nervous system and thus a higher degree of synaptic levels. The synaptic organization supports serial as well as parallel processing. Furthermore
connections from one zone to another are reciprocal and provide higher order areas a top‐down influe ce up n lower ones. n o
In humans there are six synaptic levels. The modality specific primary cortex constitutes the first synaptic level and is followed by upstream unimodal areas at the second level and downstream unimodal areas at the third level. Multimodal sensory representations converge at the fourth synaptic level in heteromodal, at the fifth level in paralimbic, and finally at the sixth level in limbic zones. The latter three areas are called transmodal areas. As it can be seen, a synaptic level gradient from the sensations first entrance into cerebral cortex, over unimodal association areas towards highly interconnected multimodal areas can be defined, see figure 1.
connect visual and auditory modalities, as can be seen by the interrupt at ring one to four ulam, 1998).
Sensory processing starts at the primary sensory cortex that forms the entrance to modality specific sensory processing. The primary sensory cortex transfers information to unimodal association areas, which are characterized through (i) their major input
Figure 1 – Primary sensory, downstream, and upstream unimodal areas for auditory (blue) and visual (green) pathways. Transmodal pathways are shown in red. Each ring shows one synaptic distance starting with one at the outer boundary and going to six at the inner circle. Nodes on the same concentric ring are reciprocally interconnected. Note that only the fifth and sixth ring inter
(Mes
from primary sensory and other unimodal association areas of that modality, (ii) constituent neurons that respond maximally to only one specific modality, and (iii) lesions that lead to behavioral deficits confined to tasks under the control of that specific sensory modality. These unimodal association areas can further be subdivided into upstream unimodal areas and downstream unimodal areas. Upstream unimodal areas encode basic features of sensation like color, motion, and form in visual extrastriate Brodmann area (BA) 18 and 19. Downstream unimodal areas encode more complex content of sensory experience like objects and faces at the inferior temporal BA 20 and 21, which functions on higher visual processing. Finally, transmodal areas bind characteristics from modality specific downstream areas and other transmodal areas and integrate these content into multimodal representations.
There are at least five higher order cognitive functions such as spatial awareness, language, memory/emotion, face‐object recognition, and working memory that are given through large‐scale distributed networks in transmodal areas. For example form the posterior parietal cortex and frontal eye field a network for spatial awareness. These networks use modality specific sensory properties to construct representations from the humans surrounding environment. The representation of a familiar person might therefore be represented in the face‐object area in the mid temporal and temporopolar cortices to remember its visual nature, in the language area in Wernicke’s and Broca’s areas to remember its name, and in memory/emotion area in the hippocampal‐
entorhinal complex to associate knowledge and emotional content to the person. These transmodal areas do not store the property information themselves but rather access them from upstream and downstream unimodal areas and serve therefore as a bibliographical node that indexes these properties (Mesulam, 1998).
Face and Object processing at the inferior temporal cortex
In the field of visual perception a lot of research has been done due on face and object recognition. The visual cues encoded in primary visual cortex (V1) are further processed in upstream and downstream unimodal areas. The downstream areas of the visual modality split into the dorsal and ventral stream of the visual pathway. The dorsal stream, located at the region of parieto‐occipital and parietal cortex, is specialized to the processing of visuospatial properties. The ventral stream, located at the middle and inferior temporal cortex, is specialized to object shape (Ungerleider & Haxby, 1994).
At the downstream region of the inferior temporal cortex within in the fusiform gyrus an important functional distinctive region was identified that is specialized to the processing of faces. This fusiform face area (FFA), that responds to a higher degree to faces than to other object categories, was first identified by Kanwisher, McDermott, and Chun (1997). Moreover it has been shown that upright faces activate this region to a higher degree than inverted ones do. The effect is known as the face inversion effect (Kanwisher, Tong, & Nakayama, 1998). The functional specialization of the FFA has been confirmed by other studies (Downing, Chan, Peelen, Dodds, & Kanwisher, 2006;
Kanwisher & Yovel, 2006; Spiridon, Fischl, & Kanwisher, 2006).
Figure 2 – Regions activated due to faces vs. objects (FFA) and scenes vs. objects (PPA) (Spiridon, et al., 2006)
Although Kanwisher and colleagues were able to show that a face specialized region in the fusiform gyrus exists, others believe that such a clear distinction cannot be drawn.
Peelen and Downing (2005) revealed that there are at least two distinctive regions at the fusiform gyrus. One region, as expected, responded to a higher degree to faces when contrasted against tools, but another nearby region responded to a higher degree to human bodies without faces when contrasted against the same control category. The authors named this body specific region the fusiform body area (FBA) with respect to the FFA. In a second experiment these authors could show that the FBA but not the FFA also responds to a higher degree to stick figure depictions of bodies than to scrambled controls, which in turn confirmed this region to encode body parts in an abstract manner (Peelen & Downing, 2005). Another group confirmed these functional distinctive regions. They conducted two localizer scans, one in standard (3mm) and one in high (1,5mm) resolution. In both scans, with a stronger selectivity for the high‐
resolution scan, the FFA and FBA was found (Schwarzlose, Baker, & Kanwisher, 2005).
The downstream area of the inferior temporal cortex contains another functional distinctive region that was identified in the parahippocampal gyrus. This region, termed the parahippocampal place area (PPA), was first described by Epstein and Kanwisher
(1998) and has since been reported as being functionally specialized to the local visual environment. It responds to passively viewed scenes, only weakly to single objects, and not at all to faces. The information about the layout of local space appears to be the critical factor for its activation and remains unaffected by the scenes complexity (Epstein & Kanwisher, 1998) and the subject’s familiarity with the scene as well as motion through the scene (Epstein, Harris, Stanley, & Kanwisher, 1999). On the other hand the PPA reflects viewpoint‐specificity, which shows that it represents the relationship between the observer and the structures that define local space (Epstein, Graham, & Downing, 2003). Powerful stimulus categories that elicit reliable activations in the PPA are landscapes and scenes as well as houses (Epstein & Kanwisher, 1998).
Other researchers have replicated the functional specialization and location of the PPA (Dow
in regions involved in visual food perception
As pictures of faces and scenes elicit activations within the downstream area of the inferior temporal cortex pictures of food should do the same. Food pictures also
ning, et al., 2006; Spiridon, et al., 2006).
Regardless of the functional specialization of the FFA for faces and the PPA for scenes it must mentioned that activity within the visual downstream area of the inferior temporal cortex is rather relative than absolute. Although neurons in the FFA respond most to faces and those in PPA most to scenes, the neuronal activity evoked in these regions is captured by other object categories as well (Ewbank, Schluppeck, & Andrews, 2005). A finding supporting this view has been revealed that different object categories evoked individual patterns of activity distributed among the ventral pathway of the visual system. The researchers were able to discriminate these object categories in a post‐hoc manner on the basis of their mean neuronal activity and even when regions were excluded that responded maximal to a specific category. Also regions that respond maximal to only one category, e.g. faces, showed discrimination patters for all categories (Haxby, et al., 2001). Another study, which elaborated the effects of adaption onto repeatedly presented inanimate objects and scenes, pointed out that adaption to inanimate objects occurred within the lateral occipital complex (LOC), the FFA, and the PPA and that adaption to places was found in the PPA and the LOC, suggesting that object representation is not restricted to those regions, which respond maximal to a single category, e.g. FFA and PPA, but rather is distributed across human visual cortex (Ewbank, et al., 2005).
Bra
constitute an object category that has to be encoded within the ventral stream of the visual pathway. The literature about food picture processing revealed activity in the left striate (Fuhrer, Zysset, & Stumvoll, 2008) and left extrastriate (Fuhrer, et al., 2008;
LaBar, et al., 2001) cortex as well as inferior visual cortex (Cornier, Von Kaenel, Bessesen, & Tregellas, 2007). Others reported visual food related activity in the fusiform gyrus within the left (Beaver, et al., 2006; Simmons, Martin, & Barsalou, 2005), the right (LaBar, et al., 2001) or both hemispheres (Cheng, Meltzoff, & Decety, 2007; Uher, Treasure, Heining, Brammer, & Campbell, 2006) and the parahippocampal gyrus in the left (St‐Onge, et al., 2005) and the right hemisphere (Beaver, et al., 2006; Cheng, et al., 2007; LaBar, et al., 2001). From this, evidence is given that pictures of food elicit neuronal activity in primary visual (striate cortex), upstream unimodal (extrastriate and inferior visual cortex), and downstream unimodal areas (FFA and PPA) and thus be trea
ort overview on research with visual food cues
When reviewing the field of neuronal food science four groups of direction can be distinguished (excluding a huge amount of work that is spend to the question of underlying neuronal causes of obesity). The first and biggest direction is formed by the study of effects of hunger and satiety on neuronal responses to visual food cues. Another
ted and processed as object categories.
Previous research into the field of food science showed a distinct pattern of the visual pathway of food object processing. In contrast to this, the control categories against which food related activations were contrasted seem to be heterogeneous. Most studies used nonfood control categories, but they consisted of food related utensils (Killgore, et al., 2003), tools (Fuhrer, et al., 2008; LaBar, et al., 2001), or were themselves of heterogeneous character (Beaver, et al., 2006; St‐Onge, et al., 2005; Uher, et al., 2006).
Others used high and low caloric pictures (Killgore & Yurgelun‐Todd, 2006) or led the subjects create mental imageries of foods (Hinton, et al., 2004). Just two studies used homogeneous controls consisting of scrambled pictures (Cheng, et al., 2007) and accordingly location pictures (Simmons, et al., 2005). The complexity of the control categories makes clear evidence about food picture related activation difficult. One has to assume that all those tools, utensils, and other nonfood objects represent a distinct object category when contrasting the distinct category of food against nonfoods. These nonfood categories do not only vary in within object features, such as eyes and mouths in faces, but also in between object features, such as the objects themselves.
A sh
field is given by the influence of reward, reward and disgust sensitivity, and affect. The third one can be made through the study of high vs. low calorie foods onto human brain activity. Only a small amount of work has attended to the question of the neuronal representation of visual food cues and food in general without comprising moderator factors. A short summary of studies related to visual food presentation is presented in table
n of the affec ive value of food (Craig, 2009; Kringelbach, 2005).
The comparison of high versus low calorie foods revealed activations in the hypothalamus (Cornier, et al., 2007; Killgore, et al., 2003), inferior temporal cortex
1.
The study of effects of hunger and satiety on food related visual stimuli showed that striate and extrastriate cortex (Fuhrer, et al., 2008), amygdala (Cheng, et al., 2007;
Hinton, et al., 2004; LaBar, et al., 2001), parahippocampal gyrus (Cheng, et al., 2007;
LaBar, et al., 2001), fusiform gyrus (LaBar, et al., 2001; Uher, et al., 2006), insula (Hinton, et al., 2004), and orbitofrontal cortex (Cheng, et al., 2007; Fuhrer, et al., 2008) activity is increased when participants are in a hungry state relative to a satiated one. These findings are in that comprehensible as this regions are associated in different aspects of food processing. The upstream extratriate cortex and downstream parahippocampal and fusiform gyrus of the visual modality are involved in object processing (Mesulam, 1998). The heteromodal orbitofrontal cortex is responsible for reward evaluation (Kringelbach, 2005). Motivation and goal directed behavior are provided through the linkage of paralimbic amygdala and heteromodal orbitofrontal cortex (Kringelbach, 2005; Mesulam, 1998). Finally, the paralimbic insula is responsible for gustatory processing (Small, et al., 1999).
Work by Gottfried, O’Doherty, and Dolan (2003) pointed out that satiety related olfactory reinforcer devaluation decreased amygdala and orbitofrontal cortex activity.
Moreover, individual’s variation in trait reward sensitivity was correlated with brain activation to appetizing foods in amygdala and orbitofrontal cortex (Beaver, et al., 2006).
Disgust sensitivity, on the other hand, was correlated with activations in the ventroanterior insula (Calder, et al., 2007). Furthermore, subject’s affective state was correlated with activations in the insula and orbitofrontal cortex (Killgore & Yurgelun‐
Todd, 2006). These studies confirm the assumption that the heteromodal orbitofrontal cortex encodes the predictive reward value of food (Kringelbach, 2005) and might, through its linkage with paralimbic insula and amygdala, also be responsible for the
evaluatio t
well as medial and dorsolateral prefrontal cortex (Killgore, et al., 2003). The activation of the hypothalamus is consistent with past findings that described this region as a center for homeostatic processes, regulation of body weight, and hunger and satiety processes (Smith & Ferguson, 2008). The activation of the heteromodal posterior parietal cortex and the increased activation of the visual downstream area to high calorie food point out that attention processes occurred during viewing of high in contrast to low calorie food. Moreover, activation of the heteromodal orbitofrontal cortex indicate that reward evaluation might be performed and prefrontal regions suggest the planning of motor actions (Mesulam, 1998).
Only two studies have focused on the neuronal representation of visual food cues.
One, conducted by Simmons, Martin, and Barsalou (2005), compared activation for location pictures with that of appetizing food pictures. They found that pictures of food activated the insula, orbitofrontal cortex, and ventral temporal cortex. The other study, carried out by St‐Onge, Sy, Heymsfield, and Hirsch (2005), explored the effects of visual and tactile presented food and nonfood items onto the neuronal processing. In a conjunction analysis, making the activation pattern independent of visual or tactile presentation, they found greater activation to food item within the parahippocampal gyrus, insula, anterior cingulate cortex, superior temporal gyrus, and hippocampus.
These two studies supported the latter findings that downstream unimodal areas in the ventral temporal cortex are bound to the object processing of visual stimuli (Mesulam, 1998) and that a linkage is given through the heteromodal orbitofrontal cortex, which is responsible for reward evaluation of seen food (Kringelbach, 2005), and paralimbic insula, which is supposed to represent the secondary gustatory cortex and thus evaluates how the food migh taste (Sm ll, et al., 1 99)t a 9 .
Taking these findings together, one can see that the visual upstream and downstream areas of the inferior temporal cortex, the heteromodal orbitofrontal cortex, and the paralimbic insula play an important role in almost every work concerning visual food presentation.
Source Modality Stimuli L/R Structures
(Beaver, et al., 2006) Visual Appetizing and nonfood object pictures1
L R R L L/R L/R
Orbitofrontal Posterior Insula
Parahippocampal gyrus Fusiform gyrus
vmPFC DLPFC (Cheng, et al., 2007) Visual Observation of grasping L Amygdala
pictures R R L/R
Parahippocampal gyrus Hypothalamus
Orbitofrontal (Cornier, et al., 2007) Visual High and low caloric
food pictures, nonfood pictures (unspecified)
L/R R L
Inferior visual cortex Hypothalamus Insula
(Fuhrer, et al., 2008) Visual Food and nonfood pictures (tools and other nonfood related objects)
L/R L L L/R
Insula Striate cortex Extrastriate cortex Cingulate cortex (Gottfried, et al., 2003) Visual
Olfactory
Arbitrary visual cues, Olfactory stimuli
L R L R L L
Orbitofrontal Anterior insula Amygdala Hypothalamus Ventral midbrain Piriform Cortex (Hinton, et al., 2004) Mental
imagery
Mental imaged foods L R L R L
Hypothalamus Amygdala Insula ACC
Orbitofrontal (Killgore & Yurgelun‐Todd,
2006)
Visual High and low caloric food pictures
R L/R L/R L/R
Lateral orbitofrontal Posterior insula Medial orbitofrontal ACC
(Killgore, et al., 2003) Visual High and low caloric food pictures, nonfood objects (food related utensils)
L L/R L R
Insula Amygdala Fusiform gyrus Medial frontal gyrus (LaBar, et al., 2001) Visual Pictures of food and
tools
R R L R L
Parahippocampal gyrus Fusiform gyrus
Amygdala Insula
Extrastriate cortex (Simmons, et al., 2005) Visual Food and location
pictures
R L L R L
Insula/Operculum Orbitofrontal ACC
Inferior temporal gyrus Fusiform gyrus
(St‐Onge, et al., 2005) Visual Tactile
Food and nonfood object pictures
R L L L/R L/R
Cingulate cortex Hippocampus
Parahippocampal gyrus Superior temporal gyrus Insula
(Uher, et al., 2006) Visual2 Food and nonfood object pictures
L/R R R L
Fusiform gyrus Lingual gyrus Angular gyrus Anterior insula
Table 1 – Selected activations found in other studies investigating food related neuronal activity (1 = as listed in supplementary table 3; 2 = gustatory activations not included; ACC = anterior cingulate cortex, vmPFC = ventromedial prefrontal cortex, DLPFC = dorsolateral prefrontal cortex).
elun‐Todd, 2006).
The consumption of food is accompanied by a reward effect that is generated and evaluated in the heteromodal orbitofrontal cortex, based on the positive outcome of its obtainment (Kringelbach, 2005). In studies related to the presentation of food pictures orbitofrontal cortex activity was observed, which indicates that a reward evaluation of food stimuli has been performed (Beaver, et al., 2006; Cheng, et al., 2007; Gottfried, et al., 2003; Hinton, et al., 2004; Killgore & Yurgelun‐Todd, 2006; Simmons, et al., 2005;
Building up a distributed netw rk for food processing o
Based on Mesulam (1998) object concepts in the brain are represented as distributed circuits of property representations across the brain’s modality specific zones. Building upon this the representation of food is at least incorporated by the properties of object shape, taste and reward value. Furthermore, viewing pictures of food should not only activate brain regions that encode the visual nature of food but moreover activate those that encode how it might taste and how rewarding it might be.
Once one part of the food processing network becomes active it may recruit the remainder structures through the process of pattern completion (Simmons, et al., 2005).
As mentioned above, visual food cues should activate brain regions that serve functions on higher visual processing, especially object encoding and recognition (Bell, Hadj‐Bouziane, Frihauf, Tootell, & Ungerleider, 2009; Mesulam, 1998). These regions are located at upstream unimodal extrastriate cortex (Fuhrer, et al., 2008; LaBar, et al., 2001) and in downstream unimodal areas at the inferior temporal cortex (Cornier, et al., 2007; Simmons, et al., 2005), especially the parahippocampal (Beaver, et al., 2006;
Cheng, et al., 2007; LaBar, et al., 2001; St‐Onge, et al., 2005) and fusiform gyrus (Beaver, et al., 2006; Cheng, et al., 2007; Killgore, et al., 2003; LaBar, et al., 2001; Simmons, et al., 2005; Uher, et al., 2006).
The paralimbic anterior insular cortex (AIC) and frontal operculum have been identified as regions that constitute the secondary gustatory cortex in humans and thus play an important role in the processing of food related gustatory properties (Sewards &
Sewards, 2001; Small, et al., 1999). The insula was activated in addition to the presentation of visual food stimuli in its frontal opercular and anterior part (Cornier, et al., 2007; Gottfried, et al., 2003; Hinton, et al., 2004; Killgore, et al., 2003; LaBar, et al., 2001; Simmons, et al., 2005; St‐Onge, et al., 2005; Uher, et al., 2006) and even some studies found a contribution of it’s posterior part (Beaver, et al., 2006; Killgore &
Yurg
Summing up, one can state that a neuronal network for food representation might at least contain the inferior temporal cortex for food object encoding and recognition, the AIC and frontal operculum for gustatory representation of food, and the orbitofrontal cortex for reward evaluation of food intake.
Motivation and salience of picture categories
Food constitutes a motivationally important stimulus category (LaBar, et al., 2001) that is highly preferable and has a salient impact on humans’ attention system, especially under the circumstances of deprivation (Stockburger, Schmaelzle, Flaisch, Bublatzky, & Schupp, 2009; Stockburger, Weike, Hamm, & Schupp, 2008). Human faces, especially with emotional content (Calvo & Nummenmaa, 2008; Eimer & Kiss, 2007), constitute a socially relevant stimulus category (Benuzzi, et al., 2007) and hence have also a salient impact that captures attention. In contrast to food and faces represents the stimulus category of houses neither a motivational important nor a social relevant category and therefore a neutral one.
Many researchers have contrasted activations caused by food pictures against those of nonfood pictures and other neutral picture categories. The contrasting of food pictures against another meaningful category, e.g. faces, has not been done nor the comparison of food related activation patterns against meaningful and neutral picture categories, as can be seen through table 1.
Influences on food perception
It has been shown that some characteristics of food stimuli are able to alter the perception of foods. For example, Walsh and colleagues have reported that children prefer such foods that are of red and green color above those of orange and yellow (Walsh, Toma, Tuveson, & Sondhi, 1990) and Blackwell (1995) found that subjects, who were instructed to rank odors in order of odor strength, which varied in odor strength and color intensity, made significantly more mistakes on judgments when the color intensity and odor strength were incompatible. Another effect was reported by Strugnell (1997), who pointed out that color (especially red) interferes with the perception of sweetness. Similar effects were found in another study, which revealed that odor and visual characteristics of seen custard affects the perception of another custards creaminess when it was ingested (de Wijk, Polet, Engelen, van Doorn, & Prinz, 2004).
Finally, taken to a larger scale, color in contrast to black and white enhances the ability
of object naming, which demonstrates it’s influence on object perception (Ostergaard &
Davidoff, 1985).
Given these findings, one might assume that brain research concerning the visual stimulation with food cues had investigated the effects of color onto the cerebral processing. However, down to the present day none researcher had published work on this.
Furthermore, each picture category consists of several low‐level stimulus characteristics such as color and brightness. For this reason, some researchers contrast their interesting picture categories against scrambled counterparts (Cheng, et al., 2007).
Deduction of thesis
Food pictures were presented to the studies participants and the signal changes evoked through this presentation in functional magnetic resonance imaging (fMRI) were contrasted against those of face and house pictures, respectively.
Taking together, with the findings reported above, blood oxygen level depend (BOLD) signal changes are expected in the visual downstream area (i) at the fusiform face area (FFA) through the contrast of face pictures against house pictures and (ii) at the parahippocampal place area (PPA) through the reverse contrast.
Moreover, food pictures should cause BOLD signal changes in the visual downstream area (iii) at the inferior temporal cortex due to object recognition and in heteromodal areas (iv) at the orbitofrontal cortex and paralimbic areas (v) at the anterior insular cortex due to the effect of pattern completion when contrasted against face and house pictures.
In addition to that the influence of color might affect neuronal responses across all contrasts and (vi) alterations in neuronal responses are expected in interaction with picture color. Furthermore, contrasting stimulus category specific activations against their scrambled counterparts (vii) should confirm the basic findings of the activation patterns stated above.
Methods
Subjects
Participants were 12 voluntary students and members of the University of Konstanz. The group contained six women and six men. No person was paid or got credit for participating. The mean age of the group was 26.83 years (SD = 6.25). The group’s mean body mass index (BMI) was 22.00 (SD = 2.42). Furthermore all persons were right handed.
Subject Age (Yrs) Sex Size (m) Weight (kg) Body Mass Index (kg/m2) Handedness
1 44 m 1,70 66 22,84 r
2 24 m 1,87 92 26,31 r
3 24 w 1,69 66 23,11 r
4 27 m 1,83 68 20,31 r
5 22 w 1,69 54 18,91 r
6 26 w 1,65 58 21,30 r
7 22 w 1,67 60 21,51 r
8 24 m 1,94 83 22,05 r
9 34 m 1,85 88 25,71 r
10 26 m 1,78 74 23,36 r
11 24 w 1,68 53 18,78 r
12 25 w 1,59 50 19,78 r
Mean 26.83 1.75 67.67 22.00
SD 6.25 0.11 13.99 2.42
Table 2 – Index of subject’s characteristics.
Stimulus Material
The stimulus material consisted of pictures with a resolution of 312 x 384 pixels and was subdivided in a set of three picture categories, namely Faces, Houses, and Food.
Each picture category included 24 different pictures. The picture category food consisted of eight meat dishes, eight vegetable dishes and eight desserts. The picture categories were available in black and white and again in color resulting in a picture set of 144 pictures. Moreover each of these pictures was divided in blocks of 3 x 3 pixels, which were rearranged in random order, resulting in 144 scrambled picture versions.
Two sets were present, one original version with black and white and colored pictures from faces, houses and food and one scrambled version of these pictures, resulting in an overall sample of 288 pictures. Examples of these stimuli can bee seen in table 19 to 21 in the appendix section.
Data acquisition
The data were acquired with a 1.5T Philips Intera magnetic resonance imaging (MRI) scanner (Philips Healthcare, Best, Netherlands). Stimuli were presented with an IBM compatible personal computer and a visual system for head coils (NordicNeuroLab, Bergen, orway). N
For functional data achievement echoplanar imaging sequences with a repetition time (TR) of 2500ms and an ascending interleaved slice acquisition order were used.
The field of view consisted of 32 slices with a resolution of 240 x 240 mm and a voxel resolution of 3 x 3 x 3.5 mm. Two scan sessions with each 290 volumes were conducted.
The structural data was acquired in T1‐weighted high‐resolution scans with 200 slices and a field of view with 256 x 256 mm and a voxel resolution of 1 x 1 x 1 mm.
Experimental Procedure
The experimental design was build up in 3 x 2 x 2 design with picture category x picture color x original/scrambled version factors.
In the run‐up to the experiment the participants were informed about potential risks of magnetic resonance imaging data acquisition and were asked about their state of health with respect to criterions for exclusion (e.g. cochlear implant, iron in body, drug intake).
Subsequent to this subjects gave their informed consent to take part in the study and filled out a questionnaire about (i) their person, (ii) handedness, (iii) former fmri studies, and (iv) potential amblyopia. In a second questionnaire participants rated their hunger state on a 9‐point Likert scale ranging from ‐4 (very hungry) to +4 (very saturated) and if true rate their appetite for something to eat on a scale ranging from 1 (little) to 10 (heavy) and name it.
Right after that the participants were prepared for data acquisition and seated in the scanner. Their feeling with the scanner environment was probed with two questions concerning valence and arousal on a 9‐point Likert scale ranging from unpleasant over neutral to pleasant for valence and from calm to excited for arousal.
Next, the first part of the stimulus presentation and functional T2‐weighted magnetic resonance imaging data acquisition was started with either all 144 colored or all 144 black and white pictures dependent on counterbalanced order. Pictures were presented in an on‐off blocked design paradigm with 12 pictures forming one on‐block
and were followed by an off‐block with a fixation‐cross in screen center. Pictures were presented for 750 ms followed by 475 ms black screen inter‐trial interval (ITI) in each on‐block. Each picture category was randomly split to from a subset of 12 pictures for blocked design presentation and contained either original or scrambled picture versions. The randomization process of the on‐block trials was constrained by that always one subset of all possible picture categories including original and scrambled versions (face‐original, house‐original, food‐original, face‐scrambled, house‐scrambled, food‐scrambled) had to be presented before the next randomization part went on. For scanner stimulus synchronization purposes each block started 15 s after the former one leaving approximately 300 ms blank in each block to adjust time lags caused from synchronization of stimulus onsets with the presentation computer’s screen frame rate.
After the first stimulus presentation part a second retrieval about the participant’s feeling with the scanner environment was done and a pause of two minutes was attached.
After this the second part of stimulus presentation and functional T2‐weighted functional magnetic resonance imaging data acquisition was started, this time with either black and white or colored pictures. Stimulus presentation and randomization parameters equaled the first part. After another two minutes pause the acquisition of the subject’s structural T1‐weighted data was performed.
After magnetic resonance imaging the participants completed a questionnaire about their eating habits (Pudel & Westenhöfer, 1989). After this, they answered three questions about (i) their concentration during the experiment, (ii) their understanding of the instructions and (iii) their potential willingness to take part in ongoing studies with food pictures. Finally, the participants were debriefed and upcoming questions were answered.
Data preprocessing and normalization
For preprocessing and normalization of functional and structural data BrainVoyager QX (Goebel, Esposito, & Formisano, 2006) was used. Raw data from the scanner output was loaded and converted into BrainVoyager’s internal FMR (functional) or VMR (structural) data format. The data were then processes to fit the needs of analysis in volume‐based space and were transformed into standardized Talairach space.
Preprocessing of functional data
Several preprocessing steps on the functional data were performed to improve data quality. To correct for time shifts between slices scanned of one volume slice scan time corrections were performed by interpolating the acquisition time of one volume’s slices to a particular time point. Head motions during scanning were corrected by spatial alignment of all volumes to the first one of the scan session by rigid‐body‐
transformations for three translation and three rotation parameters. Linear trend removal and temporal high‐pass filtering did the correction of scanner drifts. High‐pass‐
filtering was done by applying a general linear model to the voxel time courses and subtracting it from these with the residuals resulting in the filtered voxel time courses.
The design matrix consisted of sine and cosine predictors with one to three cycles per time course for the low frequencies. Spatial smoothing was done by applying a Gaussian filter with full width half maximum (FWHM) of 8mm to the volume‐based data. No spatial smoothing was used for the cortex‐based data because of later cortical alignment.
Normalization of structural data
The structural data was first corrected for intensity inhomogeneities. This was achieved by calculation of a bias field with white matter intensity changes over volume space. Next the orientation of the structural data was changed to anterior commissure (AC) ‐ posterior commissure (PC) plane by selecting the AC and PC points and converting the orientation with AC as the center point of the volume and then rotating it to be in plane and in line with the PC point. To get Talairach normalized brains the outer boundaries of the cortex were marked at anterior, posterior, superior, inferior, left and right points and then morphed to the corresponding outer boundaries of the Talairach coordinate system.
Norma ization of functional data l
To transform the functional data into Talairach space a coregistration of the functional to the corresponding structural data, using rigid‐body‐transformation, has been performed and was followed by applying all structural transformation steps to the functional data, resulting in a 4‐D Talairach normalized volume time course file (VTC), which could then be linked to the normalized structural data (VMR).
Results
The answers from the questionnaire for eating habits were combined to three scales (cognitive control, disruption of eating habits, perturbing hunger feelings), as described in the manual, for each participant and the means of these scales were calculated. For the remaining questions means were computed.
The functional data were analyzed in a random effects analysis (RFX) approach. This was done by first and second order ANOVA with computing beta weights in a first level ANOVA for the underlying predictors on single subject level for each voxel and taking these beta weights into a second ANOVA with the beta weights from the subjects being treated themselves as random subjects. Contrasts from this second level ANOVA are reported here.
Results of questionnaires
The question for hunger state revealed 1.0 (SD = 1.48) on average showing that participants were slightly saturated. Only one person reported appetite with strength of eight to a hearty meal, see table 3.
The subject’s feelings with the scanner environment were on average rated 4.92 (SD
= 1.16) for valence and 3.92 (SD = 1.56) for arousal at the beginning of the scan session followed by 5.00 (SD = 1.35) and 2.67 (SD = 0.89) at half of the scan time, respectively.
Two‐sided paired student‐t tests revealed that no statistically significant effect was present, neither for valence (t(11) = 0.723; n.s.) nor for arousal (t(11) = 0.002; n.s.).
The results for eating habits revealed on scale one for cognitive control a mean value of 9.50 (SD = 5.05), on scale two for disruption of eating habits one of 6.50 (SD = 3.40), and on scale three for perturbing hunger feelings one of 5.33 (SD = 2.81), respectively. These results showed that participants were on average in middle to high range at scale one and in middle range at scales two and three so that a normal eating behavior could be assumed. The results on subject level can be seen trough table 4.
The participant’s self reported concentration was rated at 5.25 (SD = 1.96) showing a medium concentration during the experiment, maybe because of no need to behavioral performance during the experiment. All instructions were understood well and were rated with 9.75 (SD = 0.62) and the subjects were furthermore willing to participate in ongoing studies with food pictures on a value with 9.17 (SD = 1.59), see table 4.
Subject Hunger Appetite Hunger
strength To what Feel #1 Valence
Feel #1 Arousal
Feel #2 Valence
Feel #2 Arousal
1 1 0 5 4 5 3
2 2 0 7 3 7 2
3 2 0 5 7 4 4
4 1 0 4 6 6 3
5 ‐2 1 8 Hearty meal 4 3 3 2
6 0 0 5 3 5 3
7 2 0 7 6 7 4
8 3 0 5 3 5 2
9 ‐1 0 4 3 4 3
10 2 0 3 3 3 3
11 2 0 5 2 5 1
12 0 0 5 4 6 2
Mean 1.00 4.92 3.92 5.00 2.67
SD 1.48 1.16 1.56 1.35 0.89
Table 3 – Results of subjects reported hunger feelings, if they had appetite to something or not and to what, and the feelings with the scanner environment at the first and second request on the scales arousal and valence.
Subject Cognitive
control Disruption Hunger
feelings Concentration Comprehension Future attendance
1 9 6 5 5 10 9
2 14 8 7 5 10 10
3 18 8 8 8 9 7
4 4 2 3 8 10 10
5 7 5 5 3 10 10
6 14 7 12 3 8 10
7 14 5 3 6 10 10
8 4 5 2 3 10 9
9 13 14 4 5 10 10
10 6 4 3 7 10 10
11 9 11 7 7 10 5
12 2 3 5 3 10 10
Mean 9.50 6.50 5.33 5.25 9.75 9.17
SD 5.05 3.40 2.81 1.96 0.62 1.59
Table 4 – Subjects reported eating habits on the scales for “Cognitive control”, “Disruption”, and
“Hunger feelings” as well as their concentration during the study, their comprehension of the study instructions, and their willingness to future attendance.
Results of functional data
All contrasts show t‐statistic maps with a cluster threshold of 25 voxels. Note that the functional data was resampled on the structural one and thus consists of 1 mm iso‐
voxels. Contrasts were analyzed first by picture category and further against their scrambled counterparts. These contrasts are conjunction analyses over black and white and colored pictures and allow one to study the effects independent of color. Finally, the