Abstract

In order to adequately explore the neurobiological basis of eating behavior of humans and their changes with body weight, interactions between brain areas or networks need to be investigated. In the current functional magnetic resonance imaging study, we examined the modulating effects of stimulus category (food vs. nonfood), caloric content of food, and body weight on the time course and functional connectivity of 5 brain networks by means of independent component analysis in healthy lean and overweight/obese adults. These functional networks included motor sensory, default-mode, extrastriate visual, temporal visual association, and salience networks. We found an extensive modulation elicited by food stimuli in the 2 visual and salience networks, with a dissociable pattern in the time course and functional connectivity between lean and overweight/obese subjects. Specifically, only in lean subjects, the temporal visual association network was modulated by the stimulus category and the salience network by caloric content, whereas overweight and obese subjects showed a generalized augmented response in the salience network. Furthermore, overweight/obese subjects showed changes in functional connectivity in networks important for object recognition, motivational salience, and executive control. These alterations could potentially lead to top-down deficiencies driving the overconsumption of food in the obese population.

Introduction

The development of effective weight loss interventions requires a thorough understanding of the neural basis of eating behavior of humans, which is modulated by physiological, psychological, and cognitive factors that drive and inhibit the overconsumption of food. Since we live in a nutritionally enriched environment, we are continuously exposed to food cues challenging long-term weight maintenance (Dubnov-Raz and Berry 2011). Previous neuroimaging studies imply that temporo-insulo-opercular and orbitofrontal areas are part of a network subserving food processing (e.g. Simmons et al. 2005; St-Onge et al. 2005; Porubska et al. 2006; Frank et al. 2010). This has also been confirmed by a recent meta-analysis investigating brain activation in response to viewing food compared with nonfood pictures in healthy normal-weight individuals (van der Laan et al. 2011). Among the reported brain regions, the most concurrent were the posterior fusiform gyrus, the lateral orbitofrontal cortex (OFC), and middle insula. Given that the visual system is modulated more strongly by food than by nonfood images, a greater attentional and possibly motivational salience of food objects has been proposed (LaBar et al. 2001; Killgore et al. 2003; Stockburger et al. 2008). However, there are a wide range of factors that can potentially modulate the brain response to food cues. The most frequently investigated factors are the caloric content of food images, hunger, and body weight. While viewing food pictures, the OFC and amygdala are activated more strongly during the hungry state compared with the satiated state, whereas the ventral striatum and hypothalamus are modulated more strongly by high- versus low-caloric food cues (van der Laan et al. 2011). Furthermore, several studies have shown that body weight is an important factor that influences the response of the brain to food images. In the hungry state, obese subjects showed an increased activation in the anterior cingulate cortex (ACC) and medial prefrontal cortex (MPFC) while viewing food pictures (Martin et al. 2009). Furthermore, brain regions related to hedonic aspects of food cues, especially the striatum, showed significantly reduced dopamine D2 receptor availability (Wang et al. 2001) and exhibited greater activation to high-caloric food stimuli in obese individuals (Rothemund et al. 2007; Stoeckel et al. 2008). Additionally, an interaction between caloric content and level of hunger has been observed (Goldstone et al. 2009; Siep et al. 2009; Frank et al. 2010). These studies exemplify that the regulation of eating behavior is a complex process that cannot be assigned to isolated brain areas. In order to adequately explore the neural basis involved in the eating behavior of humans and their changes with body weight, we need to study interactions between brain areas or brain networks. A recent research has shown that different behavioral domains, such as visual perception, working memory, action and somesthesis, or executive function, localize to dissociable functional brain networks, identified during resting state as well as during activation conditions (Smith et al. 2009; Laird et al. 2011). Changes in functional networks have also been related to several different medical conditions such as schizophrenia (Meda et al. 2009), depression (Greicius et al. 2007), and Alzheimer's disease (Greicius et al. 2004). Additionally, we recently discovered that obese subjects showed prominent differences in resting-state networks (Kullmann et al. 2012). In the current functional magnetic resonance imaging (fMRI) study, we analyzed task-based functional brain networks in lean and overweight/obese subjects using a one-back visual recognition task with food and nonfood stimuli, as described previously (Frank et al. 2010). The aim of our study was to examine the modulating effects of stimulus category (food vs. nonfood), caloric content (high vs. low), and body weight (lean vs. overweight/obese subjects) on the time course and functional connectivity of brain networks identified using independent component analysis (ICA). We hypothesized that networks of the visual system are modulated by the stimulus category, whereas caloric content and body weight will modulate networks involved in higher cognitive processing.

Materials and Methods

Subjects and Behavioral Tasks

Twenty-four right-handed subjects participated in the study after an overnight fast of at least 10 h: 12 lean [6 women, body mass index (BMI) range 19.4–22.5 kg/m2, and age range 22–29 years], 3 overweight, and 9 obese subjects (6 women, BMI range 28.4–34.4 kg/m2, and age range 21–28 years). Lean subjects were required to have a BMI of 18.5–24.0 kg/m2. Overweight and obese subjects were required to have a BMI >25 kg/m2. Overweight and obese subjects collectively will be referred to as “overweight/obese.” To minimize circadian influence, subjects were scanned between 7 a.m and 8 a.m. Before the experiment, subjects confirmed their fasting state and rated their subjective feeling of hunger on a visual analog scale from 0 to 10 (0, not hungry at all and 10, very hungry). In addition, blood glucose and plasma insulin levels before the experiment were within the fasting range for all subjects. Prior to the experiment, subjects filled out a short German version of the “Profile of Mood States” (POMS; McNair et al. 1981), which is a 35-item questionnaire that evaluates depression, fatigue, anger, and vigor.

Before the measurement day, all subjects underwent a medical screening to assure that they did not suffer from psychiatric, neurological, or metabolic diseases. To address psychiatric diseases, the Patient Health Questionnaire (Löwe et al. 2002) was used; to assess eating behavior, subjects took the German Three-Factor Eating Questionnaire (TFEQ) (Pudel and Westenhöfer 1989). Any volunteer treated for chronic disease or taking any kind of medication other than oral contraceptives was excluded at screening. All subjects were normal-sighted or had corrected-to-normal vision. Informed written consent was obtained from all subjects, and the protocol was approved by the local Ethics Committee of the Medical Faculty of the University of Tübingen. The demographic characteristics of the subjects are shown in Table 1.

Table 1

Subjects' characteristics in lean versus overweight/obese group

 Lean group Overweight/obese group P-value 
Gender (female/male) 6/6 6/6 – 
Age (years) 22.91 ± 2.10 24.66 ± 2.42 0.07 
BMI (kg/m221.16 ± 1.13 30.46 ± 1.77 <0.001 
Fasting glucose (mmol/L) 4.94 ± 0.34 4.96 ± 0.41 0.9 
Fasting insulin (pmol/L) 37 ± 15 77 ± 33 0.0014 
Hunger rating 4.97 ± 2.41 2.82 ± 2.18 0.032 
TFEQ    
 Restraint eating 4.75 ± 4.55 9.33 ± 4.7 0.024 
 Disinhibition 3.91 ± 1.24 6.08 ± 3.57 0.06 
 Hunger 5.66 ± 3.14 4.50 ± 2.31 0.321 
Accuracy (%)    
 One-back visual recognition task 87 ± 0.07 88 ± 0.08 0.442 
Reaction time (s)    
 One-back visual recognition task 0.84 ± 0.16 0.73 ± 0.09 0.08 
 Lean group Overweight/obese group P-value 
Gender (female/male) 6/6 6/6 – 
Age (years) 22.91 ± 2.10 24.66 ± 2.42 0.07 
BMI (kg/m221.16 ± 1.13 30.46 ± 1.77 <0.001 
Fasting glucose (mmol/L) 4.94 ± 0.34 4.96 ± 0.41 0.9 
Fasting insulin (pmol/L) 37 ± 15 77 ± 33 0.0014 
Hunger rating 4.97 ± 2.41 2.82 ± 2.18 0.032 
TFEQ    
 Restraint eating 4.75 ± 4.55 9.33 ± 4.7 0.024 
 Disinhibition 3.91 ± 1.24 6.08 ± 3.57 0.06 
 Hunger 5.66 ± 3.14 4.50 ± 2.31 0.321 
Accuracy (%)    
 One-back visual recognition task 87 ± 0.07 88 ± 0.08 0.442 
Reaction time (s)    
 One-back visual recognition task 0.84 ± 0.16 0.73 ± 0.09 0.08 

Note: Data represent mean ± SD. Comparison of unadjusted loge-transformed data by ANOVA.

We employed a one-back visual recognition task as described previously (Frank et al. 2010). A stimulus set of 48 color food- (high and low caloric) and 48 nonfood-related pictures matched for physical complexity, arousal, valence, and appetite was used. Nonfood images comprised objects that had no association with eating. Pictures were presented either during a one-back visual recognition task or during a null-back control task. For the one-back visual recognition task, subjects had to press a button with the right index finger at each picture depending on whether the presented picture was the same or different from the one presented before. For the control task, subjects had to press a button as soon as they saw the picture. The experiment consisted of 2 sessions, lasting 7 min each. Each session comprised 12 different stimulation blocks, resulting in 4 blocks for the control task (food and nonfood pictures combined), 4 blocks for the one-back visual recognition task with food pictures, and 4 blocks with nonfood pictures. The food picture blocks were further divided into blocks with high- and low-caloric food pictures. Before each block, subjects were informed whether the upcoming task is a control task or a one-back task. Stimuli were presented for 1.5 s with an interstimulus interval varying randomly between 2 and 3 s. Before and after each block, a blank screen with a red fixation cross in the middle was shown for 14 s. Presentation software (version 10.2, www.neurobs.com) was used for stimulus presentation and response data collection via button press. Subjects were instructed to look at the middle of the picture or the fixation cross that would appear between stimulus presentations.

Imaging Protocol

Whole-brain fMRI data were obtained by using a 3.0 T scanner (Siemens Tim Trio, Erlangen, Germany). Each session consisted of 226 scans (repetition time = 2 s, echo time = 30 ms, matrix 64 × 64, flip angle 90°, voxel size 3 × 3 × 3.75 mm3, 30 slices), and the images were acquired in ascending order. In addition, high-resolution T1-weighted anatomical images (magnetization-prepared rapid gradient echo: 176 slices, matrix 256 × 224, 1 × 1 × 1 mm3) of the brain were obtained. During the scan procedure, subjects were lying in a supine position. Their heads were stabilized by foam padding around the head within a head coil to minimize head movements. Pictures were projected on a tilted mirror mounted on the head coil.

Spatial and Temporal Preprocessing

Preprocessing and statistical analysis of the fMRI data were performed using SPM5 (Wellcome Trust Centre for Neuroimaging, London, UK). Images were realigned and resliced to the first image after unwarping in the phase-encoding direction (anterior–posterior) to account for susceptibility by movement artifacts. A mean image was created and coregistered to the T1 structural image. The anatomical image was normalized to Montreal Neurological Institute (MNI) space, and the resulting parameter file was used to normalize the functional images (voxel size 3 × 3 × 3 mm3). Finally, the normalized images were smoothed with a 3-dimensional isotropic Gaussian kernel [full width at half maximum (FWHM): 12 mm]. fMRI data were highpass- (cutoff period 128 s) and lowpass-filtered [autoregression model AR(1)]. The residual motion correction plots showed that all subjects had less than 1 mm or 0.5° movement.

Independent Component Analysis

Group spatial ICA (Calhoun et al. 2001) was used to decompose the smoothed, normalized fMRI images into 40 components using the GIFT software (http://icatb.sourceforge.net/) as follows. Dimension, to determine the number of components, was estimated using the minimum description length criterion, modified to account for spatial correlation (Li et al. 2007). Data from all subjects were then concatenated, and this aggregate data set was reduced to 40 temporal dimensions using principal component analysis, followed by an independent component (IC) estimation using infomax algorithm (Bell and Sejnowski 1995). A time course for each component and its corresponding spatial map, which represent a real contribution to this component time course, was obtained. Finally, for each subject, spatial maps were then reconstructed and converted to z-values. The z-values of the spatial maps represent the fit of a specific voxel blood oxygen level-dependent time course to the group-averaged component's time course.

Component Selection

We first discarded noise-related components that showed spatial correlation r2 > 0.02 for white matter or r2 > 0.05 for cerebrospinal fluid. One of the strengths of the ICA is the possibility to identify noise-related components caused by head and eye movements, ventricle activity, and other signal artifacts (McKeown et al. 2003). This approach has been employed by other fMRI studies using ICA (Jafri et al. 2008; Kim et al. 2009). We also ran group ICA 50 times using ICASSO (Himberg et al. 2004) and used the cluster quality index (Iq) to validate ICA reliability. In order to examine the relationship between the ICs and the experimental paradigm, a regression was performed on the ICA time courses with the GLM design matrix (in SPM5), which contained 4 regressors: high-caloric food, low-caloric food, and nonfood for the visual recognition task, and one regressor for the control task. This is referred to as temporal sorting in GIFT. The regression analysis resulted in a set of beta-weights, which represent the degree of synchrony between the component and reference time course, indicating the extent of modulation of the network in the task condition (Meda et al. 2009). We averaged the beta-weights across the 2 sessions for each subject each for high-caloric food, low-caloric food, nonfood, and control. Positive and negative beta-weights indicate positive and negative correlations, respectively, for each condition, which means that the network was either positively or negatively modulated by the task. The resulting components were sorted according to their R2 statistic. We selected those components that were among the top 2 rank-ordered components for high-caloric, low-caloric, nonfood, and control conditions (Supplementary Table S1). Furthermore, we selected the default-mode network (DMN) and the salience network a priori by using spatial sorting in GIFT, applying masks derived from the Wake Forest University Pick Atlas (http://www.fmri.wfubmc.edu/download.htm). For the DMN mask, we used precuneus, posterior cingulate, and Brodmann's area (BA) 7, 10, and 39 (Correa et al. 2007). For the salience network mask, we used the ACC and bilateral insula cortex (Dosenbach et al. 2006, 2007; Laird et al. 2011).

Statistical Analysis of Functional Networks

The beta-weights, generated from the temporal-sorting procedure, of 4 positively modulated networks and 1 negatively modulated network were entered into a repeated-measures analysis of variance (ANOVA) (within-subject factor with 3 levels: food, nonfood, and control conditions; between-subject factor: body weight; covariate: age). To analyze category-specific effects within and between groups, paired t-tests for the contrast food versus nonfood and high- versus low-caloric condition were calculated (P < 0.05, Bonferroni-corrected) (SPSS version 19; SPSS Inc., Chicago, IL, USA).

Spatial maps generated by the ICA reflect functional connectivity maps within each IC. To determine the associated regions of these maps and to identify group differences, a full factorial model was calculated in SPM5 (within factor: session; between factor: body weight; covariate: age). The resulting statistical maps were masked with the component-specific group map to explore results within this network only. All group tests were thresholded at P < 0.05 [family-wise error (FWE)-corrected].

Functional Network Connectivity (FNC) Analysis

Following the ICA, we further explored the modulating effects of body weight on the temporal relationship between ICA time courses, which is a measure of FNC (Jafri et al. 2008). The analysis was performed using the FNC toolbox available from http://www.ece.unm.edu/~vcalhoun/mialab/Software. The FNC approach examines this specific temporal correlation in a pairwise manner using a maximal lagged correlation approach, in which, in our case, 15 pairwise combinations were computed using the 6 selected components taking 2 at a time. Correlation values were calculated for each subject. To determine which components were significantly correlated, we used a 1-sample t-test to assess group-level significance for each of the 15 pairwise combinations in the lean and overweight/obese subjects separately (P < 0.05, corrected for multiple comparisons). We then used a 2-sample t-test to determine statistically significant differences in correlation between groups. For a summary sketch of the functional data analysis, see Supplementary Figure S1.

Blood Sampling and Analysis

Blood glucose was measured by a HemoCue blood glucose photometer using the glucose dehydrogenase method (HemoCueAB, Aengelholm, Sweden). Plasma insulin levels were determined using commercial chemiluminescence assays for ADVIA Centaur (Siemens Medical Solutions, Fernwald, Germany).

Voxel-Based Morphometry

T1-weighted images were processed and examined using the VBM8 toolbox with default parameters (http://dbm.neuro.uni-jena.de/vbm.html) implemented in the SPM8 software (Wellcome Department of Imaging Neuroscience Group; http://www.fil.ion.ucl.ac.uk./spm). Images were bias-corrected, tissue-classified, and registered using linear and nonlinear transformations, within a unified model (Ashburner and Friston 2005). Gray matter (GM) segments were multiplied by the nonlinear components derived from the normalization matrix in order to preserve actual GM values locally (modulated GM volumes) and not multiplied by the linear components of the registration in order to account for individual differences in brain orientation, alignment, and size globally. Finally, the modulated volumes were smoothed with a Gaussian kernel of 10 mm FWHM.

Voxel-wise GM differences between lean and overweight/obese subjects were examined by means of a 2-sample t-test, using age and gender as confounding covariates. In order to avoid possible edge effects between tissue types, we excluded all voxels with GM values less than 0.1 (absolute threshold masking). Results were thresholded at P < 0.05 (FWE-corrected).

Behavioral Data

The analysis of the behavioral data [hunger rating, reaction time (RT), and accuracy] was performed using SPSS (version 19; SPSS Inc.). A one-way ANOVA was used for the hunger rating. For RT and accuracy, a repeated-measures ANOVA was applied (within-subject factor with 4 levels: high-caloric, low-caloric, nonfood, and control conditions and between-subject factor: body weight). Correlation analyses were calculated between beta-weights and behavioral measurements (i.e. TFEQ scores and subjective hunger ratings) corrected for BMI by using partial correlation analysis. A P-value less than 0.05 was considered to be statistically significant.

Results

Behavioral Data

Accuracy

There was no significant difference in accuracy between lean and overweight/obese subjects and conditions. The average accuracy was 88% (SD = 0.08) for the one-back visual recognition task.

Reaction Time

A repeated-measures ANOVA of the RT showed a main effect of condition (F3,66 = 164.653, P < 0.001). In the control task, both lean and overweight/obese subjects were significantly faster (mean = 0.54 ± 0.13 s) compared with the one-back visual recognition task (mean = 0.79 ± 0.14 s). No main effect of body weight (F1,22 = 3.362, P = 0.08) or interaction (F3,66 = 1.464, P = 0.232) was observed.

Hunger Rating

The ratings for the subjective feeling of hunger showed significant differences between lean and overweight/obese subjects (one-way ANOVA: P = 0.03). Lean subjects were significantly more hungry (mean = 4.9 ± 2.41) than overweight/obese subjects (mean = 2.82 ± 2.18).

Questionnaires

There was no statistically significant difference in mood states (i.e. depression, fatigue, anger, and vigor) assessed by POMS between lean and overweight/obese subjects on the day of measurement (P > 0.05). The TFEQ revealed a significant difference between lean and overweight/obese subjects for the restraint eating scale (P = 0.024) and a trend for disinhibition (P = 0.06) (Table 1).

Component Identification

By means of ICA, 40 ICs were extracted for the overweight/obese group as well as for the lean group. After discarding noise-related components, 22 potentially functional relevant components were extracted for both groups. A temporal correlation analysis revealed 3 ICs that were significantly positively modulated by the visual recognition and control tasks for food and nonfood stimuli (Supplementary Table S1). These 3 ICs comprised the motor sensory, extrastriate visual, and the temporal visual association system. Therefore, we refer to these networks as motor sensory network, extrastriate visual network, and temporal visual association network (Laird et al. 2011). The motor sensory network included activation of the precentral and postcentral gyri bilaterally (more activation on the left side), left supplementary motor area, and right cerebellum (Fig. 1A). The extrastriate visual network included the cuneus, calcarine gyrus (BA 18), and inferior parietal cortex (BA 40) (Fig. 1B). The temporal visual association network included the inferior occipital gyrus (BA 19) and the fusiform gyrus (BA 37) (Fig. 1C). On the basis of prior knowledge, 2 further components were selected by using a priori masks: the salience network and DMN, which were positively and negatively modulated by the task, respectively. The salience network included the medial frontal gyrus (BA 9), anterior and middle cingulate cortices (BA 32 and BA 24), insular cortex, and inferior parietal cortex (BA 40) (Fig. 1D). The DMN included activation of the precuneus, posterior cingulate, and BA 7, 10, and 39 (Fig. 1E).

Figure 1.

(A) Group-averaged motor sensory network, (B) extrastriate visual network, (C) temporal visual association network, (D) salience network, and (E) DMN extracted from fMRI data of lean and overweight/obese subjects, thresholded at P < 0.05 (FWE-corrected).

Figure 1.

(A) Group-averaged motor sensory network, (B) extrastriate visual network, (C) temporal visual association network, (D) salience network, and (E) DMN extracted from fMRI data of lean and overweight/obese subjects, thresholded at P < 0.05 (FWE-corrected).

All 5 components showed an Iq greater than 0.9 (Fig. 1).

Main Effect of Body Weight on Functional Connectivity

In the extrastriate visual network, the left inferior occipital gyrus showed decreased functional connectivity strength in the overweight/obese group compared with the lean group (P < 0.05, FWE-corrected) (Fig. 2). In the salience network, the ACC and the superior medial frontal gyrus (BA 9), part of the MPFC, showed increased functional connectivity strength in the overweight/obese group compared with the lean group (P < 0.05, FWE-corrected) (Fig. 2). In the DMN, the precuneus (BA 7) showed increased functional connectivity strength in the overweight/obese group compared with the lean group (P < 0.05, FWE-corrected) (Table 2).

Table 2

Main effect of body weight

Location Hemisphere Brodmann's area MNI coordinates
 
z-value 
X Y Z 
Lean minus overweight/obese       
 Extrastriate visual network       
  Inferior occipital gyrus 18 −6 −102 −3 6.31 
Overweight/obese minus lean       
 Salience network       
  Superior medial frontal cortex −6 33 33 4.93 
  Anterior cingulate cortex R/L 32 ±6 45 4.38 
 Default mode network       
  Precuneus R/L −78 33 5.37 
   −12 −84 30 4.15 
Location Hemisphere Brodmann's area MNI coordinates
 
z-value 
X Y Z 
Lean minus overweight/obese       
 Extrastriate visual network       
  Inferior occipital gyrus 18 −6 −102 −3 6.31 
Overweight/obese minus lean       
 Salience network       
  Superior medial frontal cortex −6 33 33 4.93 
  Anterior cingulate cortex R/L 32 ±6 45 4.38 
 Default mode network       
  Precuneus R/L −78 33 5.37 
   −12 −84 30 4.15 

Note: Significant difference between lean and obese subjects in the extrastriate visual association, salience, and DMNs (P < 0.05, FWE-corrected).

Figure 2.

Altered functional connectivity in overweight/obese subjects in the extrastriate visual and salience networks. Color map represents significant (P < 0.001, uncorrected for display) voxels of decreased functional connectivity in blue and increased functional connectivity in orange. Color bar represents t-values. Overweight/obese compared with lean subjects showed decreased functional connectivity in the calcarine gyrus (blue) in the extrastriate visual network and increased functional connectivity in the superior medial frontal gyrus and ACC (orange) in the salience network.

Figure 2.

Altered functional connectivity in overweight/obese subjects in the extrastriate visual and salience networks. Color map represents significant (P < 0.001, uncorrected for display) voxels of decreased functional connectivity in blue and increased functional connectivity in orange. Color bar represents t-values. Overweight/obese compared with lean subjects showed decreased functional connectivity in the calcarine gyrus (blue) in the extrastriate visual network and increased functional connectivity in the superior medial frontal gyrus and ACC (orange) in the salience network.

Functional Network Connectivity

Even though individual temporally defined components derived through ICA are spatially independent, they can still be significantly temporally correlated with each other; therefore, we analyzed FNC. Out of the 10 possible combinations, we observed 8 significant correlation/connectivity combinations between the 5 selected ICs, with the strongest correlations for the visual networks (data not shown). The combination between the temporal visual association and salience networks showed a significant difference between lean and overweight/obese subjects, whereas overweight/obese subjects revealed increased FNC [t(22) = −2.19, P = 0.03].

Modulation of Functional Networks

A repeated-measures ANOVA of the temporal beta-weights showed no main effect of condition for all 5 networks (P > 0.05). A main effect of body weight was observed only in the salience network (P = 0.02). A significant condition-by-body-weight interaction was observed in the temporal visual association network (P = 0.02) and extrastriate visual network (P = 0.01). The category-specific analysis showed significant within-group effects for food versus nonfood conditions in the extrastriate visual network for the lean and overweight/obese groups (P < 0.001) (Fig. 3A) and in the temporal visual association network for lean subjects only (P = 0.01) (Fig. 3B). Furthermore, the category-specific analysis revealed a significant difference between high- and low-caloric conditions in the extrastriate visual network for the lean and overweight/obese groups (P < 0.05) (Fig. 3A) and in the salience network for lean subjects only (P = 0.03) (Fig. 3C). The category-specific analysis also revealed between-group effects in the extrastriate visual network for the low-caloric condition (P = 0.05), in the temporal visual association network for high- and low-caloric conditions (P = 0.01 and P = 0.02, respectively), and in the salience network for the low-caloric condition (P = 0.003) (Fig. 3AC). Category-specific results are summarized in Table 3.

Table 3

Modulation of functional networks

Independent components (ICs) Between-group category-specific effects
 
Within-group category-specific effects
 
Hi Lo NF F versus NF
 
Hi versus lo
 
F versus NF
 
Hi versus lo
 
Lean
 
Obese
 
Extrastriate visual network t = 1.79, n.s. t = 2.11, P = 0.05 t = 0.216, n.s. t = 8.74, P < 0.001 t = 2.52, P = 0.02 t = 5.34, P < 0.001 t = 2.99, P = 0.01 
Temporal visual association network t = 2.96, P = 0.01 t = 2.63, P = 0.02 t = 0.791, n.s. t = 2.98, P = 0.01 n.s. n.s. n.s. 
Salience network t = −1.14, n.s. t = −3.84, P = 0.003 t = −1.56, n.s. n.s. t = 2.47, P = 0.03 n.s. n.s. 
Independent components (ICs) Between-group category-specific effects
 
Within-group category-specific effects
 
Hi Lo NF F versus NF
 
Hi versus lo
 
F versus NF
 
Hi versus lo
 
Lean
 
Obese
 
Extrastriate visual network t = 1.79, n.s. t = 2.11, P = 0.05 t = 0.216, n.s. t = 8.74, P < 0.001 t = 2.52, P = 0.02 t = 5.34, P < 0.001 t = 2.99, P = 0.01 
Temporal visual association network t = 2.96, P = 0.01 t = 2.63, P = 0.02 t = 0.791, n.s. t = 2.98, P = 0.01 n.s. n.s. n.s. 
Salience network t = −1.14, n.s. t = −3.84, P = 0.003 t = −1.56, n.s. n.s. t = 2.47, P = 0.03 n.s. n.s. 

Note: Beta-weights, generated from the temporal regression analysis, were entered into paired t-tests to identify category-specific effects within and between groups for food (F) versus nonfood (NF) and high (hi) versus low caloric (lo) task conditions. n.s., not significant.

Figure 3.

Modulation of functional networks. This shows the degree to which the component time course was associated with the presented conditions during the visual recognition task for lean and overweight/obese subjects (±SE). (A) For the “extrastriate visual network,” within-group comparisons revealed a significant difference between food and nonfood conditions and between high- and low-caloric conditions in lean and overweight/obese subjects. Between-group comparisons revealed significant differences between lean and overweight/obese subjects for the low-caloric condition. (B) For the “temporal visual association network,” within-group comparisons revealed a significant difference between food and nonfood conditions in lean subjects only. Between-group comparisons revealed significant differences between lean and overweight/obese subjects for the high- and low-caloric conditions. (C) For the “salience network,” within-group comparisons showed no significant difference between food and nonfood conditions in lean and overweight/obese subjects. However, in lean subjects, there was a significant difference for high- versus low-caloric conditions. Between-group comparisons revealed significant differences between lean and overweight/obese subjects for the low-caloric condition. (D) Significant negative correlation between temporal network modulations of food stimuli and restraint eating. *P < 0.05 and **P < 0.001.

Figure 3.

Modulation of functional networks. This shows the degree to which the component time course was associated with the presented conditions during the visual recognition task for lean and overweight/obese subjects (±SE). (A) For the “extrastriate visual network,” within-group comparisons revealed a significant difference between food and nonfood conditions and between high- and low-caloric conditions in lean and overweight/obese subjects. Between-group comparisons revealed significant differences between lean and overweight/obese subjects for the low-caloric condition. (B) For the “temporal visual association network,” within-group comparisons revealed a significant difference between food and nonfood conditions in lean subjects only. Between-group comparisons revealed significant differences between lean and overweight/obese subjects for the high- and low-caloric conditions. (C) For the “salience network,” within-group comparisons showed no significant difference between food and nonfood conditions in lean and overweight/obese subjects. However, in lean subjects, there was a significant difference for high- versus low-caloric conditions. Between-group comparisons revealed significant differences between lean and overweight/obese subjects for the low-caloric condition. (D) Significant negative correlation between temporal network modulations of food stimuli and restraint eating. *P < 0.05 and **P < 0.001.

Correlation of Temporal Beta-Weights with Behavioral Measurements

We found significant correlations between subjective hunger ratings and functional network modulation for all conditions of the one-back visual recognition task, adjusted for BMI. For the extrastriate visual network, we found a significant positive correlation with high-caloric food condition (P = 0.009, R2 = 0.24). For the temporal visual association network, we found significant positive correlations with high-caloric food (P = 0.04, R2 = 0.14), with low-caloric food (P = 0.02, R2 = 0.16), and with nonfood conditions (P = 0.001, R2 = 0.34). For the salience network, we found a significant negative correlation with low-caloric food conditions (P = 0.01, r2 = 0.18). Furthermore, we found a significant negative correlation between TFEQ restraint eating scale and temporal visual association network modulation for food stimuli corrected for BMI (r = −0.477, P = 0.021) (Fig. 3D).

Structural Differences Between Lean and Overweight/Obese Subjects

To assess possible causes of altered functional connectivity, we analyzed our data for structural differences between both groups. The voxel-wise analysis revealed no significant GM differences at a threshold of P < 0.05 (FWE-corrected).

Discussion

Thus far, functional brain networks underlying eating behavior of humans and their changes with body weight have not been sufficiently explored. In this fMRI study, we examined the modulating effects of stimulus category, caloric content, and body weight on the time course and functional connectivity of 5 brain networks. Subjects did not have to explicitly evaluate food stimuli, but a one-back task was used to keep their attention focussed. We found an extensive modulation elicited by food stimuli in the 2 visual networks and the salience network, with a dissociable pattern in the time course and functional connectivity between lean and overweight/obese subjects. Subjective feeling of hunger strongly modulated the above-mentioned networks for all conditions of the one-back visual recognition task, independent of BMI; furthermore, we observed a significant relationship between the restraint eating scale and temporal visual association network modulation for food stimuli. A direct group comparison of the spatial maps showed altered functional connectivity in overweight/obese subjects within the inferior occipital gyrus, ACC, MPFC, and precuneus. Furthermore, overweight/obese subjects showed increased FNC between the temporal visual association and salience network. Regions of the affected functional network did not differ significantly in volumetric aspects between the 2 groups, confirming the functional nature of the observed alterations.

The 2 functional networks of the visual system, together with the motor sensory network, showed the highest degree of synchrony with the one-back visual recognition task. The extrastriate visual network, including primary and secondary visual areas, was modulated by the stimulus category showing higher correlations for food than for nonfood conditions, even though we matched the images for complexity, valence, and arousal. This is in line with a recent magnetoencephalography study, in which we could show a food versus nonfood effect in primary visual areas as early as 120 ms after the stimulus onset, indicating a categorization effect in the primary visual cortex (Stingl et al. 2010). Furthermore, the extrastriate visual network was also influenced by the caloric content; hereby high-caloric foods elicited the strongest response. Albeit most neuroimaging studies have focussed on subcortical brain structures or the frontal cortex to be modulated by the caloric content (Rothemund et al. 2007; Stoeckel et al. 2008), recent research has also shown the importance of the visual system in discriminating high- from low-caloric foods. The lateral occipital cortices were identified to track energy value of food images in a recent electroencephalography study (Toepel et al. 2009), whereas fMRI studies have identified the fusiform gyrus to elicit an enhanced response to high-caloric foods (Killgore and Yurgelun-Todd 2007; Siep et al. 2009; Frank et al. 2010). On a behavioral level, an attentional bias toward food pictures was observed in women, which was enhanced by body weight and hunger (Nijs et al. 2010); furthermore, food items with higher energy values were more rapidly processed, leading to a faster response to high-caloric foods compared with low-caloric foods (Harrar et al. 2011). Considering the fact that subjects were in a fasting state in this study, high-caloric foods probably had the highest incentive value, leading to heightened visual processing in the extrastriate visual network. The temporal visual association network, in contrast, was substantially affected by body weight. This network included higher visual processing areas as the fusiform gyrus. Here only in lean subjects, a modulation by the stimulus category was revealed leading to an augmented response to food cues, whereas overweight/obese subjects showed no modulation according to the stimulus category and a general reduced response to food stimuli compared with lean subjects. Furthermore, the modulation of the temporal visual association network for food stimuli was highly influenced by the eating-related behavioral assessment “restraint eating,” which was significantly higher in the overweight/obese group compared with the lean group. Restraint eating is frequently initiated as a reaction to weight gain as a cognitive-mediated effort to eat less than desired (Johnson et al. 2011). In our study, a decreased modulation of the visual system by food stimuli was observed in subjects with high restraint eating, indicating a possible attempt to reduce visual attention to food in order to eat less. Therefore, our results imply that increased body weight can alter neural processing in high-level visual areas, which is given further merit by the fact that we found decreased functional connectivity in the occipital cortex. Since we found no effect of BMI on primary visual areas, we suggest that the lean versus overweight/obese differences are not driven exclusively by bottom-up deficiencies in sensory processing. Possible top-down regulation of visual processing mediated by the frontal cortex and the influence of body weight will be discussed subsequently.

During visual processing, it is assumed that goal-directed activity modulation is achieved via connections between control regions in the prefrontal and visual cortices (Miller and D'Esposito 2005). The salience network, also referred to as “core network” in several studies, is consistently recruited by cognitively demanding tasks suggested to control goal-directed behavior (Dosenbach et al. 2006, 2007; Seeley et al. 2007; Laird et al. 2011). In our paradigm, we found significant correlations between the time course of the task and the salience network with stronger associations for the overweight/obese group. However, only in lean subjects, a modulation by the caloric content was observed with an intensified response to high-caloric foods, whereas overweight/obese subjects showed a generalized augmented response to food stimuli without modulation by the caloric content. Furthermore, the salience network has been proposed to be a translational network linking cognition and emotion (Laird et al. 2011). The ACC and anterior insula respond to personal salience including motivational, emotional, cognitive, or homeostatic across tasks through underlying interoceptive autonomic processing integrating highly processed sensory data (Bush et al. 2000; Critchley 2005), whereas the prefrontal and parietal cortices are exceedingly involved in executive control operating on identified salience by directing attention toward relevant stimuli (Seeley et al. 2007). Previous studies have also frequently reported the MPFC and ACC to show greater activity for high- than for low-caloric foods (Killgore et al. 2003; Siep et al. 2009) in the hungry state, exhibiting enhanced activity in overweight/obese compared with lean subjects (Stoeckel et al. 2008; Martin et al. 2009). Concomitantly, we observed differences between lean and overweight/obese subjects in functional connectivity, leading to higher functional connectivity in the ACC and superior medial frontal cortex within the salience network and increased FNC between the salience and temporal visual association networks in the overweight/obese group. Therefore, we propose the salience network as a mediator of top-down regulation, thereby leading to extensive visual processing for salient stimuli, especially increasing awareness for food in the fasted state. Our findings imply that overweight/obese subjects might use cognitive strategies, such as restraint eating or disinhibition, to overcome the strong personal salience to food cues, which could potentially lead to top-down deficiency.

Another potential factor influencing the cerebral response to food cues is the subjective feeling of hunger. Even though in our study all subjects were in a fasted state, we found a significant difference in hunger ratings between lean and overweight/obese subjects. This could be due to social desirability, leading to lower ratings for overweight/obese subjects. Indeed, hunger significantly correlated with food and nonfood conditions for 3 functional brain networks, independent of BMI. The extrastriate visual network was positively modulated by hunger for the high-caloric food conditions, whereas the salience network was negatively modulated for the low-caloric food conditions. The temporal visual association network, however, was modulated by hunger for all conditions including nonfood cues, indicating that hunger can lead to a generalized enhanced visual attention. Hence, further studies are needed to elucidate functional network differences by manipulating the state of hunger/satiation to discriminate between the effect of body weight and hunger.

The DMN and the motor sensory network were not modulated by the stimulus category, caloric content, body weight, or hunger. The DMN is active when subjects are in an awake, resting phase without any task, but its activity diminishes during specific goal-directed behaviors (Raichle et al. 2001). We found differences in DMN functional connectivity between lean and overweight/obese subjects, revealing increased functional connectivity in the precuneus in the overweight/obese group, which converges with a previous study in which we found increased functional connectivity in the precuneus during resting state in obese subjects (Kullmann et al. 2012). Functional connectivity has been linked to task difficulty (Esposito et al. 2009); hence, we can speculate that overweight/obese subjects subjectively needed more resources to perform the one-back task. The engagement of the motor sensory network can be explained by fact that subjects had to press 2 different buttons during the task. Nevertheless, this network was not modulated by the stimulus category, caloric content, body weight, or hunger.

Finally, we acknowledge some limitations of the present study. First, the number of subjects investigated was rather small. Secondly, eating behavior and the neural response to food cues can potentially be influenced by the menstrual cycle. Yet, we did not assess the cycle phase of our female subjects. Therefore, further studies are needed to evaluate FNC during one consistent phase of the menstrual cycle to rule out possible confounding hormonal effects.

Conclusion

In the present study, we identified functional brain networks important for food processing. We found an extensive visual processing of food stimuli in 2 visual networks, which can be explained by the motivational salience of the stimuli leading to heightened attention. The visual networks and the salience network revealed a dissociable pattern in the time course and functional connectivity between lean and overweight/obese subjects. Changes in functional connectivity in networks important for object recognition, motivational salience, and executive control were observed in overweight/obese subjects. We propose that these differences could potentially lead to top-down deficiencies driving the overconsumption of food, which might be a cause or consequence of obesity and an important factor for the development of new weight loss strategies.

Supplementary Material

Supplementary material can be found at: http://www.cercor.oxfordjournals.org/.

Funding

This study was supported by the “Kompetenznetz Adipositas” (Competence Network for Adiposity) (FKZ: 01GI0837 and 01GI0849) funded by the German Federal Ministry of Education and Research (BMBF) and the German Center for Diabetes Research (DZD e.V.).

Notes

We thank Margarete Bayer, Maike Borutta, and Anna Bury for their excellent technical assistance. Conflict of Interest: None declared.

References

Ashburner
J
Friston
KJ
Unified segmentation
NeuroImage
 , 
2005
, vol. 
26
 (pg. 
839
-
851
)
Bell
AJ
Sejnowski
TJ
An information-maximization approach to blind separation and blind deconvolution
Neural Comput
 , 
1995
, vol. 
7
 (pg. 
1129
-
1159
)
Bush
G
Luu
P
Posner
MI
Cognitive and emotional influences in anterior cingulate cortex
Trends Cogn Sci
 , 
2000
, vol. 
4
 (pg. 
215
-
222
)
Calhoun
VD
Adali
T
Pearlson
GD
Pekar
JJ
A method for making group inferences from functional MRI data using independent component analysis
Hum Brain Mapp
 , 
2001
, vol. 
14
 (pg. 
140
-
151
)
Correa
N
Adali
T
Calhoun
VD
Performance of blind source separation algorithms for fMRI analysis using a group ICA method
Magn Reson Imaging
 , 
2007
, vol. 
25
 (pg. 
684
-
694
)
Critchley
HD
Neural mechanisms of autonomic, affective, and cognitive integration
J Comp Neurol
 , 
2005
, vol. 
493
 (pg. 
154
-
166
)
Dosenbach
NU
Fair
DA
Miezin
FM
Cohen
AL
Wenger
KK
Dosenbach
RA
Fox
MD
Snyder
AZ
Vincent
JL
Raichle
ME
, et al.  . 
Distinct brain networks for adaptive and stable task control in humans
Proc Natl Acad Sci USA
 , 
2007
, vol. 
104
 (pg. 
11073
-
11078
)
Dosenbach
NU
Visscher
KM
Palmer
ED
Miezin
FM
Wenger
KK
Kang
HC
Burgund
ED
Grimes
AL
Schlaggar
BL
Petersen
SE
A core system for the implementation of task sets
Neuron
 , 
2006
, vol. 
50
 (pg. 
799
-
812
)
Dubnov-Raz
G
Berry
EM
The dietary treatment of obesity
Med Clin North Am
 , 
2011
, vol. 
95
 (pg. 
939
-
952
)
Esposito
F
Aragri
A
Latorre
V
Popolizio
T
Scarabino
T
Cirillo
S
Marciano
E
Tedeschi
G
Di Salle
F
Does the default-mode functional connectivity of the brain correlate with working-memory performances?
Arch Ital Biol
 , 
2009
, vol. 
147
 (pg. 
11
-
20
)
Frank
S
Laharnar
N
Kullmann
S
Veit
R
Canova
C
Hegner
YL
Fritsche
A
Preissl
H
Processing of food pictures: influence of hunger, gender and calorie content
Brain Res
 , 
2010
, vol. 
1350
 (pg. 
159
-
166
)
Goldstone
AP
Prechtl de Hernandez
CG
Beaver
JD
Muhammed
K
Croese
C
Bell
G
Durighel
G
Hughes
E
Waldman
AD
Frost
G
, et al.  . 
Fasting biases brain reward systems towards high-calorie foods
Eur J Neurosci
 , 
2009
, vol. 
30
 (pg. 
1625
-
1635
)
Greicius
MD
Flores
BH
Menon
V
Glover
GH
Solvason
HB
Kenna
H
Reiss
AL
Schatzberg
AF
Resting-state functional connectivity in major depression: abnormally increased contributions from subgenual cingulate cortex and thalamus
Biol Psychiatr
 , 
2007
, vol. 
62
 (pg. 
429
-
437
)
Greicius
MD
Srivastava
G
Reiss
AL
Menon
V
Default-mode network activity distinguishes Alzheimer's disease from healthy aging: evidence from functional MRI
Proc Natl Acad Sci USA
 , 
2004
, vol. 
101
 (pg. 
4637
-
4642
)
Harrar
V
Toepel
U
Murray
MM
Spence
C
Food's visually perceived fat content affects discrimination speed in an orthogonal spatial task
Exp Brain Res
 , 
2011
, vol. 
214
 (pg. 
351
-
356
)
Himberg
J
Hyvarinen
A
Esposito
F
Validating the independent components of neuroimaging time series via clustering and visualization
NeuroImage
 , 
2004
, vol. 
22
 (pg. 
1214
-
1222
)
Jafri
MJ
Pearlson
GD
Stevens
M
Calhoun
VD
A method for functional network connectivity among spatially independent resting-state components in schizophrenia
NeuroImage
 , 
2008
, vol. 
39
 (pg. 
1666
-
1681
)
Johnson
F
Pratt
M
Wardle
J
Dietary restraint and self-regulation in eating behavior
Int J Obes (Lond). doi: 10.138/ijo.2011.156
 , 
2011
Killgore
WD
Young
AD
Femia
LA
Bogorodzki
P
Rogowska
J
Yurgelun-Todd
DA
Cortical and limbic activation during viewing of high- versus low-calorie foods
NeuroImage
 , 
2003
, vol. 
19
 (pg. 
1381
-
1394
)
Killgore
WD
Yurgelun-Todd
DA
Positive affect modulates activity in the visual cortex to images of high calorie foods
Int J Neurosci
 , 
2007
, vol. 
117
 (pg. 
643
-
653
)
Kim
DI
Manoach
DS
Mathalon
DH
Turner
JA
Mannell
M
Brown
GG
Ford
JM
Gollub
RL
White
T
Wible
C
, et al.  . 
Dysregulation of working memory and default-mode networks in schizophrenia using independent component analysis, an fBIRN and MCIC study
Hum Brain Mapp
 , 
2009
, vol. 
30
 (pg. 
3795
-
3811
)
Kullmann
S
Heni
M
Veit
R
Ketterer
C
Schick
F
Haring
HU
Fritsche
A
Preissl
H
The obese brain: association of body mass index and insulin sensitivity with resting state network functional connectivity
Hum Brain Mapp
 , 
2012
, vol. 
33
 (pg. 
1052
-
1061
)
LaBar
KS
Gitelman
DR
Parrish
TB
Kim
YH
Nobre
AC
Mesulam
MM
Hunger selectively modulates corticolimbic activation to food stimuli in humans
Behav Neurosci
 , 
2001
, vol. 
115
 (pg. 
493
-
500
)
Laird
AR
Fox
PM
Eickhoff
SB
Turner
JA
Ray
KL
McKay
DR
Glahn
DC
Beckmann
CF
Smith
SM
Fox
PT
Behavioral interpretations of intrinsic connectivity networks
J Cogn Neurosci
 , 
2011
, vol. 
23
 (pg. 
4022
-
4037
)
Li
YO
Adali
T
Calhoun
VD
Estimating the number of independent components for functional magnetic resonance imaging data
Hum Brain Mapp
 , 
2007
, vol. 
28
 (pg. 
1251
-
1266
)
Löwe
B
Spitzer
RL
Zipfel
S
Herzog
W
PHQ-D. Gesundheitsfragebogen für Patienten
 , 
2002
Karlsruhe
Pfizer
Martin
LE
Holsen
LM
Chambers
RJ
Bruce
AS
Brooks
WM
Zarcone
JR
Butler
MG
Savage
CR
Neural mechanisms associated with food motivation in obese and healthy weight adults
Obesity (Silver Spring)
 , 
2009
, vol. 
18
 (pg. 
254
-
260
)
McKeown
MJ
Hansen
LK
Sejnowsk
TJ
Independent component analysis of functional MRI: what is signal and what is noise?
Curr Opin Neurobiol
 , 
2003
, vol. 
13
 (pg. 
620
-
629
)
McNair
DM
Lorr
M
Droppelmann
LF
Manual for the Profile of Mood States
 , 
1981
San Diego (CA)
Educational and Industrial Testing Service
Meda
SA
Stevens
MC
Folley
BS
Calhoun
VD
Pearlson
GD
Evidence for anomalous network connectivity during working memory encoding in schizophrenia: an ICA based analysis
PLoS One
 , 
2009
, vol. 
4
 pg. 
e7911
 
Miller
BT
D'Esposito
M
Searching for “the top” in top-down control
Neuron
 , 
2005
, vol. 
48
 (pg. 
535
-
538
)
Nijs
IM
Muris
P
Euser
AS
Franken
IH
Differences in attention to food and food intake between overweight/obese and normal-weight females under conditions of hunger and satiety
Appetite
 , 
2010
, vol. 
54
 (pg. 
243
-
254
)
Porubska
K
Veit
R
Preissl
H
Fritsche
A
Birbaumer
N
Subjective feeling of appetite modulates brain activity: an fMRI study
NeuroImage
 , 
2006
, vol. 
32
 (pg. 
1273
-
1280
)
Pudel
D
Westenhöfer
J
Fragebogen zum Eßverhalten (FEV). Handanweisung
 , 
1989
Göttingen
Hogrefe
Raichle
ME
MacLeod
AM
Snyder
AZ
Powers
WJ
Gusnard
DA
Shulman
GL
A default mode of brain function
Proc Natl Acad Sci USA
 , 
2001
, vol. 
98
 (pg. 
676
-
682
)
Rothemund
Y
Preuschhof
C
Bohner
G
Bauknecht
HC
Klingebiel
R
Flor
H
Klapp
BF
Differential activation of the dorsal striatum by high-calorie visual food stimuli in obese individuals
NeuroImage
 , 
2007
, vol. 
37
 (pg. 
410
-
421
)
Seeley
WW
Menon
V
Schatzberg
AF
Keller
J
Glover
GH
Kenna
H
Reiss
AL
Greicius
MD
Dissociable intrinsic connectivity networks for salience processing and executive control
J Neurosci
 , 
2007
, vol. 
27
 (pg. 
2349
-
2356
)
Siep
N
Roefs
A
Roebroeck
A
Havermans
R
Bonte
ML
Jansen
A
Hunger is the best spice: an fMRI study of the effects of attention, hunger and calorie content on food reward processing in the amygdala and orbitofrontal cortex
Behav Brain Res
 , 
2009
, vol. 
198
 (pg. 
149
-
158
)
Simmons
WK
Martin
A
Barsalou
LW
Pictures of appetizing foods activate gustatory cortices for taste and reward
Cereb Cortex
 , 
2005
, vol. 
15
 (pg. 
1602
-
1608
)
Smith
SM
Fox
PT
Miller
KL
Glahn
DC
Fox
PM
Mackay
CE
Filippini
N
Watkins
KE
Toro
R
Laird
AR
, et al.  . 
Correspondence of the brain's functional architecture during activation and rest
Proc Natl Acad Sci USA
 , 
2009
, vol. 
106
 (pg. 
13040
-
13045
)
Stingl
KT
Rogic
M
Stingl
K
Canova
C
Tschritter
O
Braun
C
Fritsche
A
Preissl
H
The temporal sequence of magnetic brain activity for food categorization and memorization—an exploratory study
NeuroImage
 , 
2010
, vol. 
52
 (pg. 
1584
-
1591
)
Stockburger
J
Weike
AI
Hamm
AO
Schupp
HT
Deprivation selectively modulates brain potentials to food pictures
Behav Neurosci
 , 
2008
, vol. 
122
 (pg. 
936
-
942
)
Stoeckel
LE
Weller
RE
Cook
EW
Twieg
DB
Knowlton
RC
Cox
JE
Widespread reward-system activation in obese women in response to pictures of high-calorie foods
NeuroImage
 , 
2008
, vol. 
41
 (pg. 
636
-
647
)
St-Onge
MP
Sy
M
Heymsfield
SB
Hirsch
J
Human cortical specialization for food: a functional magnetic resonance imaging investigation
J Nutr
 , 
2005
, vol. 
135
 (pg. 
1014
-
1018
)
Toepel
U
Knebel
JF
Hudry
J
le Coutre
J
Murray
MM
The brain tracks the energetic value in food images
NeuroImage
 , 
2009
, vol. 
44
 (pg. 
967
-
974
)
van der Laan
LN
de Ridder
DT
Viergever
MA
Smeets
PA
The first taste is always with the eyes: a meta-analysis on the neural correlates of processing visual food cues
NeuroImage
 , 
2011
, vol. 
55
 (pg. 
296
-
303
)
Wang
GJ
Volkow
ND
Logan
J
Pappas
NR
Wong
CT
Zhu
W
Netusil
N
Fowler
JS
Brain dopamine and obesity
Lancet
 , 
2001
, vol. 
357
 (pg. 
354
-
357
)