Principles of functional MRI

fMRI has revolutionised the research of functional brain mapping and is now the most widely used method for mapping brain activity. fMRI measures brain activity indirectly through the haemodynamic changes associated with neural activation, and typically uses the BOLD contrast. BOLD contrast was introduced in the early 1990s⁽ Reference Ogawa, Lee and Kay ^{1 )}, and relies on the different magnetic properties of oxygenated and deoxygenated blood, which are diamagnetic and paramagnetic, respectively⁽ Reference Okada ^{2 )}. Paramagnetic deoxyhaemoglobin leads to a local magnetic field distortion within and around vessels, causing the protons to dephase, reducing the T ₂/T ₂* relaxation time and the MR signal compared with oxyhaemoglobin.

On brain activity, neural demand increases leading to an increase in glucose and oxygen consumption with an increase in both CBF and cerebral blood volume to deliver these nutrients. However, the increase in CBF, and hence the supply of oxygenated blood, increases to a level greater than that required by demand (over compensation). This leads to a local increase in blood oxygenation in active brain regions, resulting in a local increase in T ₂/T ₂* relaxation time and thus a small, but detectable, increase in MR image intensity termed the BOLD response⁽ Reference Thulborn, Waterton and Matthews ^{3 )}, Fig. 1. This effect can be observed using T ₂* or T ₂ weighted imaging sequences using gradient-echo or spin-echo schemes, respectively. Although gradient-echo echo planar imaging is most typically used, the change in BOLD MR signal intensity has a characteristic response, termed the haemodynamic response function, with the increase in MR signal typically delayed by 6 s following stimulus onset and having a 1–5 % signal change at a magnetic field strength of 3 T.

Fig. 1. Schematic representation of (a) the haemodynamic changes which lead to a blood oxygenation level-dependent (BOLD) signal, and (b) the BOLD haemodynamic response function.

Since BOLD-fMRI is a non-invasive technique, it can be repeatedly used. The spatial resolution is typically 3 ml isotropic, but higher spatial resolution can be achieved at increasing magnetic field strength, such as at an ultra-high field (UHF) of 7 T. The main limitation of the BOLD technique arises from the vascular origin of the signal, which inherently limits its temporal resolution and quantitation⁽ Reference Buxton, Uludag and Dubowitz ^{4 ,} Reference Hoge, Atkinson and Gill ^{5 )}. In addition to developing stronger magnetic fields to enhance the spatial resolution of the BOLD response and brain mapping, measuring the direct change in CBF due to neuronal activation using arterial spin labelling provides absolute quantification of the signal change and hence neuronal activity.

In arterial spin labelling, arterial blood water is magnetically labelled before it reaches the tissue of interest by applying a single radiofrequency pulse⁽ Reference Edelman, Siewert and Darby ^{6 ,} Reference Kim ^{7 )} or train of pulses⁽ Reference Wu, Fernandez-Seara and Detre ^{8 )}. After a period of time (post-label delay), the labelled blood ‘endogenous tracer’ flows into the slice of interest, where it exchanges with tissue water. During this time, an image is taken (called the tag or label image). In this image, the blood water is in a different magnetisation state from that of the static tissue water. In addition to the label image, a control image is acquired without labelling of inflowing arterial blood. Label and control images are then subtracted to yield a perfusion weighted image that can be quantified in terms of cerebral blood flow in ml/100 g per min⁽ Reference Buxton, Wong and Frank ^{9 )}. Although directly measuring CBF using arterial spin labelling yields a better spatial correlation with the actual site of the active brain region than BOLD, its temporal resolution is lower than BOLD due to the time required to collect both a label and control pair. In addition, the signal-to-noise ratio and contrast-to-noise ratio are inherently lower than BOLD⁽ Reference Silva, Lee and Yang ^{10 ,} Reference Duong, Kim and Ugurbil ^{11 )}; however increasing the field strength increases both of these measures due to longer longitudinal relaxation time (T ₁) at higher magnetic field strength.

Functional MRI paradigm designs to study nutrition

In fMRI, various designs and approaches are used for studying the mechanism underlying food intake and appetite. One of the most direct methods is the presentation of taste, aroma or flavour stimuli or food images in a block design, where stimuli are presented for a period of time followed by a rest period. In an fMRI study, a set of BOLD images covering the whole brain (a brain volume) is typically collected each 2–3 s during the presentation of stimuli, and to increase the sensitivity to detect active brain areas, hundreds of brain volumes are acquired during many repeats of a stimulus. This leads to an fMRI paradigm typically taking 10–20 min. Signal averaging and statistical processing are required due to the difficulty in detecting the low BOLD signal changes against background physiological noise signal fluctuations⁽ Reference Purdon and Weisskoff ^{12 )}. However, BOLD sensitivity increases supra-linearly with magnetic strength, one reason for the current demand for UHF MR scanners.

Alternatively, the brains’ response to food or food cues can be measured under different physiological conditions, such as the fasted state v. fed state, or the presence of exogenously administered or endogenously produced gut hormones. Another approach is to study the resting state signal fluctuations in the brain, where individuals do not perform any tasks during scanning (this is termed resting state fMRI). Subjects are scanned under different physiological conditions and the connectivity between different brain regions is studied. CBF-fMRI studies provide an alternative assessment of the effect of physiological state by measuring the absolute cerebral blood flow at different time points, which can then be compared with a baseline measure of CBF.

Functional MRI responses to taste, aroma and oral somatosensation

The sensory properties of food and their hedonic effects are important drivers of food intake. The use of fMRI in recent years has improved the understanding of the cortical representation of sensory properties of food in human subjects. Although taste stimulation has consistently activated the same brain regions across multiple fMRI studies, including insula and overlaying operculum, the transduction mechanisms remain incompletely characterised. Previous studies reported the primary taste cortex is located within the anterior insula/frontal operculum⁽ Reference Small, Zald and Jones-Gotman ^{13 –} Reference Veldhuizen, Albrecht and Zelano ^{15 )} with secondary projections to the orbitofrontal cortex (OFC)⁽ Reference Francis, Rolls and Bowtell ^{16 )}, amygdala⁽ Reference O'Doherty, Rolls and Francis ^{17 )}, anterior cingulate cortex (ACC)⁽ Reference Small, Gregory and Mak ^{18 )}, ventral striatum⁽ Reference O'Doherty, Dayan and Friston ^{19 )} and dorsolateral prefrontal cortex⁽ Reference Kringelbach, de Araujo and Rolls ^{20 )}. O'Doherty et al. were the first to investigate the cortical response to pleasant (glucose) and aversive (salt) taste stimuli by assessing stimuli against a tasteless control stimulus⁽ Reference O'Doherty, Rolls and Francis ^{21 )}. In individual subject analysis, the OFC showed separate areas to be activated in response to these two tastes. Small et al. also highlighted activations in the OFC to sweet and aversive taste, and investigated the neural response to taste intensity when valence was held constant, showing activation of the amygdala and mid-insula⁽ Reference Small, Gregory and Mak ^{18 )}. Furthermore, the amygdala responded to both sweet and aversive taste (when intensity was held constant), with preferential activation to sweet taste, providing evidence that the amygdala is not solely involved in processing aversive taste. Other activations including ACC have been reported in many fMRI studies⁽ Reference Francis, Rolls and Bowtell ^{16 ,} Reference Small, Gregory and Mak ^{18 ,} Reference O'Doherty, Rolls and Francis ^{21 ,} Reference de Araujo, Kringelbach and Rolls ^{22 )}, while the rolandic operculum of the parietal cortex has been identified as a part of the primary gustatory cortex⁽ Reference Ogawa, Wakita and Hasegawa ^{23 ,} Reference Faurion, Cerf and Van De Moortele ^{24 )}. A recent meta-analysis confirmed widespread activation of the insula and operculum in response to taste stimuli, diverging from posterior to the most anterior junction of insula and operculum, in addition to activation in medial OFC, pregenual ACC and mediodorsal thalamus⁽ Reference Veldhuizen, Albrecht and Zelano ^{15 )}.

In response to aroma stimuli, fMRI studies report robust activation in higher-order olfactory areas, including the OFC, insula and ACC⁽ Reference Francis, Rolls and Bowtell ^{16 ,} Reference O'Doherty, Rolls and Francis ^{25 ,} Reference Savic, Gulyas and Larsson ^{26 )}. In contrast, the primary olfactory cortex has shown no⁽ Reference Yousem, Williams and Howard ^{27 ,} Reference Fulbright, Skudlarski and Lacadie ^{28 )} or inconsistent activations in some studies⁽ Reference O'Doherty, Rolls and Francis ^{25 ,} Reference Weismann, Yousry and Heuberger ^{29 )}. This may be due to its small structure, susceptibility artefacts in this region or habituation effects⁽ Reference Poellinger, Thomas and Lio ^{30 )}.

Oral somatosensation plays a crucial role in many aspects of flavour perception. fMRI studies have revealed brain regions involved in tactile components such as astringency⁽ Reference Cerf-Ducastel, Van de Moortele and MacLeod ^{31 )}, burn⁽ Reference Rudenga, Green and Nachtigal ^{32 )} and temperature⁽ Reference Guest, Grabenhorst and Essick ^{33 )}. Textural properties of food are known to correlate with fat content, and fat may also act as a chemical stimulus, with NEFA stimulating taste receptor cells for the detection of fat⁽ Reference Gilbertson ^{34 ,} Reference Mattes ^{35 )}. Very few studies have investigated the cortical representation to oral viscosity and oral fat in the mouth. de Araujo and Rolls⁽ Reference de Araujo and Rolls ^{36 )} and De Celis Alonso et al. ⁽ Reference De Celis Alonso, Marciani and Head ^{37 )} investigated the representation of oral viscosity (carboxymethyl cellulose) and manugel alginate gel, respectively. Both studies showed activation in the mid- and anterior insula, with the mid-insula representing the somatosensory properties of the oral activity (texture), other areas, including post-central gyrus and rolandic/parietal operculum were also reported⁽ Reference De Celis Alonso, Marciani and Head ^{37 )}. In addition, the sensory and hedonic aspects of oral fat have been investigated in a number of studies using BOLD and CBF measures. Human responses to oral fat have been found to activate taste, texture and reward areas⁽ Reference de Araujo and Rolls ^{36 ,} Reference Eldeghaidy, Marciani and McGlone ^{38 –} Reference Frank, Linder and Kullmann ^{40 )}. Responses to oral fat in the form of pure fat have been found in the ACC extending to the OFC and hypothalamus, with the rostral ACC and medial OFC being activated independent of viscosity, leading to the suggestion that it is these areas that process the hedonic properties of fat⁽ Reference de Araujo and Rolls ^{36 )}. Moreover, the subjective rating of texture and flavour pleasantness of oral fat was found to correlate with activity in the mid-OFC and ACC, areas thought to represent flavour pleasantness⁽ Reference Grabenhorst, Rolls and Parris ^{39 )}. This study also suggested that the pregenual cingulate cortex is activated by converging flavour and fat signals, hence representing overall pleasantness. Further investigations on the effect of fat content on the cortical response show a positive correlation with increasing fat level/concentration in taste, texture and reward areas⁽ Reference Eldeghaidy, Marciani and McGlone ^{38 –} Reference Frank, Linder and Kullmann ^{40 )}. Most recently, the somatosensory cortex has been shown to be involved in the processing of oral flavoured fat through functional coupling with the OFC, for a high-fat pleasant sample compared with a low-fat food with the same flavour. It has been suggested that this may reflect the role of somatosensory cortex in the processing of pleasant-flavoured oral fat⁽ Reference Grabenhorst and Rolls ^{41 )}. Understanding the cortical encoding of the perceived response to fat is important in relation to the pleasantness of food and appetite control, and it provides additional and valuable information to the food industry in order to improve the design of products that are lower in fat but still rewarding to eat, and hence decrease the risk of obesity and associated complications.

Flavour perception of foods or beverages is a complex multimodal process, engaging taste, olfaction and texture modalities, and does not result from the simple addition of these stimuli⁽ Reference Small, Jones-Gotman and Zatorre ^{14 )}. Despite taste, aroma and oral somatosenation systems being anatomically dissociated at the peripheral level, they are highly integrated at the cortical level, and the exact mechanisms that lead to flavour perception are still not well understood. Small⁽ Reference Small ^{42 )} has suggested a model for flavour perception, which postulates that the oral somatomotor areas play the principal role in binding the taste, olfactory and associated tactile sensations into the unitary flavour percept. fMRI provides a great tool to understand the cortical association and integration between the different modalities. Evidence from recent fMRI studies show overlapping brain responses to taste and aroma in areas including anterior insula, ACC, OFC and amygdala⁽ Reference Small, Voss and Mak ^{43 –} Reference McCabe and Rolls ^{47 )}. Verhagen and Engelen performed a meta-analysis, which suggested the anterior insula to be a multi-modal area activated by both taste and aroma stimuli⁽ Reference Verhagen and Engelen ^{48 )}. However, the mechanism of interaction is poorly understood. In addition, the cortical representation to oral somatosensation and taste was shown to overlap in the insula and operculum⁽ Reference Cerf-Ducastel, Van de Moortele and MacLeod ^{31 –} Reference Guest, Grabenhorst and Essick ^{33 )}. Although fMRI provides a powerful method in revealing the cortical response to flavour perception, much is still to be discovered in flavour perception.

Phenotype and its modulation of functional MRI responses

The perception of taste is known to vary widely across individuals. There are many factors that contribute to an individual's taste perception and subsequent food preferences and energy intake, including the density of taste papillae on the tongue and genetic differences in taste receptors⁽ Reference Bartoshuk, Duffy and Miller ^{49 )}. The most well-researched taste phenotyping is the perception of the bitterness taste. In behavioural studies, this can be assessed by measuring a subjects’ sensitivity to a bitter chemical called 6-n-propylthiouracil (PROP). Subjects are asked to taste a filter paper soaked in super saturated PROP solution and rate for bitterness intensity on a Generalised Labelled Magnitude Scale⁽ Reference Bartoshuk, Duffy and Fast ^{50 )}. Non-tasters cannot taste the bitterness of PROP (about 25 % of the population), medium tasters (about 50 %) sense the bitterness while accepting it and super-tasters (about 25 %) find the taste of PROP unacceptable. Tasters (medium- and super-tasters) are more sensitive than non-tasters to the bitterness of caffeine, and to the sweetness of sucrose⁽ Reference Chang, Chung and Kim ^{51 )}, some artificial sweeteners⁽ Reference Bartoshuk, Duffy and Miller ^{49 )} and salt taste⁽ Reference Miller and Reedy ^{52 )}. Moreover, some studies have also shown that PROP tasters are also more sensitive to fat⁽ Reference Duffy, Bartoshuk and Lucchina ^{53 )}, oral irritation⁽ Reference Duffy, Peterson and Bartoshuk ^{54 )} and oral temperature⁽ Reference Manrique and Zald ^{55 )}. Variability in PROP taster status is, in part, explained by the TAS2R receptor gene family⁽ Reference Duffy, Davidson and Kidd ^{56 )}. Previous research showed that the density of fungiform papillae is significantly higher in PROP super-tasters⁽ Reference Duffy, Davidson and Kidd ^{56 ,} Reference Duffy and Bartoshuk ^{57 )} suggesting that this could be the source of increased sensitivity to taste stimuli and oral sensitivity. Essick et al. showed a strong correlation between PROP tasters and oral somatosensation⁽ Reference Essick, Chopra and Guest ^{58 )}, with supertasters having improved lingual tactile spatial acuity (to embossed letters) compared with medium tasters or non-tasters.

Recently, a new taste phenotype known as thermal taster status has been described⁽ Reference Cruz and Green ^{59 )}. Thermal stimulation of small areas of the tongue was found to elicit a ‘phantom’ taste in some individuals. Thermal sourness and/or saltiness can be perceived by cooling the tongue, whereas re-warming the tongue from an initial cooling period can evoke a sweetness taste. In addition, thermal tasters were found to be more sensitive than thermal non-tasters to taste and olfactory stimuli⁽ Reference Green and George ^{60 )}, and oral stimulation including temperature⁽ Reference Bajec and Pickering ^{61 )}. The mechanism underlying this is still under investigation, and appears to not be associated with fungiform papillae density, as no correlation has been found previously. Fungiform papillae are surrounded and sometimes innervated by the trigeminal nerve, which carries information from the tongue related to chemical and physical properties of foods, and this may lead to a correlation between PROP tasters and thermal tasters. Recent studies also show there is a relationship between the trigeminal system and PROP taster status⁽ Reference Bajec and Pickering ^{61 )}.

Very few studies have investigated the impact of taste phenotype on the primary gustatory cortex and oral somatosensory areas. Eldeghaidy et al. showed that the cortical response to oral fat in PROP tasters was highly correlated in key taste, texture, and reward processing areas, with a significant increase in BOLD response with PROP taster status (super-taster>taster>non-taster)⁽ Reference Eldeghaidy, Marciani and McGlone ^{38 )}, Fig. 2. More recently, Clark et al., showed that the cortical response to taste and carbonation in thermal tasters was significantly higher in several cortical areas involved in flavour processing compared with thermal non-tasters⁽ Reference Clark, Francis and Bealin-Kelly ^{62 )}. Furthermore, the variation in genotype, such as the α-synuclein gene and TaqIA A1, has also been shown to be associated with the BOLD response to taste stimuli in the OFC and striatum⁽ Reference Stice, Spoor and Bohon ^{63 ,} Reference Felsted, Ren and Chouinard-Decorte ^{64 )} and regions known to be rich in dopaminergic axon terminals, including ACC, amygdala, thalamus and putamen⁽ Reference Wilcox, Claus and Blaine ^{65 )}. These results highlight new information about the cortical features of different phenotypes, and emphasise the importance of phenotyping participants for fMRI taste studies to increase the statistical power of group analysis.

Fig. 2. (colour online) Cortical response to oral fat. (a) Blood oxygenation level-dependent response to oral fat delivery, shown with false discovery rate (FDR) corrected at P < 0·05, overlaid on a T ₁-weighted image. (b) Region of interest analysis of the T-scores showing a significant increase in the oral somatosensory (SI, SII and mid-insula) and reward area (amygdala) in 6-n-propylthiouracil supertasters⁽ Reference Eldeghaidy, Marciani and McGlone ^{38 )}. ACC, anterior cingulate cortex.

Effect of hunger and satiety on functional MRI response

A number of fMRI studies have investigated the cortical modulation to satiation. Most studies show increased activation in insula, amygdala, OFC parahippocampal cortex, thalamus, hypothalamus, ACC and caudate in a hunger/fasting state, compared with increased activation in prefrontal cortex and inferior parietal lobe in a satiated/fed state (areas that show decreased activation in fasting/hunger areas). Viewing food images when hungry modulates the response in appetite-related brain regions, including the amygdala, parahippocampal gyrus and lateral OFC, compared with the fed state⁽ Reference Fuhrer, Zysset and Stumvoll ^{66 –} Reference van der Laan, de Ridder and Viergever ^{69 )}. In addition, manipulating food motivation by varying energy content of food images is associated with fMRI activation in hypothalamus, ventral stratium, cerebellum and frontal middle gyrus to high-energy food images compared with low-energy images⁽ Reference Fuhrer, Zysset and Stumvoll ^{66 ,} Reference van der Laan, de Ridder and Viergever ^{69 –} Reference Goldstone, de Hernandez and Beaver ^{71 )}.

In addition to visual stimulation with food images, studies using oral food cues have also shown modulation of brain responses by satiation. Key brain areas attenuated due to satiation include the insula, amygdala, striatum and OFC. Suppression in OFC, amygdala, ventral striatum is reported in many studies and may reflect the decrease in pleasantness and desire to eat⁽ Reference O'Doherty, Rolls and Francis ^{25 ,} Reference Smeets, Graaf and Stafleu ^{72 ,} Reference Smeets, Weijzen and de Graaf ^{73 )}. Suppression in the anterior insula activity on satiation suggests its role in feeding behaviour in addition to its taste sensory processing⁽ Reference de Araujo, Kringelbach and Rolls ^{22 ,} Reference O'Doherty, Rolls and Francis ^{25 ,} Reference Small, Zatorre and Dagher ^{74 –} Reference Gautier, Del Parigi and Chen ^{76 )}. The hypothalamus is widely recognised as the gatekeeper to control food intake, highly influenced by nutrients, with evidence that hypothalamic dysfunction may lead to obesity⁽ Reference Thaler and Schwartz ^{77 )}. Prior studies have shown the hypothalamus is modulated by satiety⁽ Reference Gautier, Del Parigi and Chen ^{76 ,} Reference Smeets, de Graaf and Stafleu ^{78 )}; however this modulation is inconsistent across BOLD studies, and may be due to its small structure, variability in position and close proximity to the sinus cavity. Recent CBF studies show a decrease in activation of the hypothalamus after the consumption of glucose⁽ Reference Page, Chan and Arora ^{79 )} and a high-fat meal⁽ Reference Frank, Linder and Kullmann ^{40 )} compared with baseline. In addition, the decrease in hypothalamus CBF correlated positively with the insula gustatory brain region⁽ Reference Frank, Linder and Kullmann ^{40 )}, indicating the interaction between homeostatic and taste areas in response to a high-fat meal.

How do hormonal responses influence the cortex?

Food intake and its termination are affected by internal and external factors. The homeostatic system controls appetite and food intake via anorectic and appetitive hormones connecting to the neural circuitry, and converging through the hypothalamus to stimulate or inhibit food intake⁽ Reference von Ruesten, Steffen and Floegel ^{80 )}. The non-homeostatic mechanisms including food palatability (through sight, smell and taste sensory), habitual, sociocultural, emotional and economic influences also drive food intake and appetite. It is now widely accepted that, there is extensive cross-modulation between homeostatic and non-homeostatic systems, with regard to appetite control. Fig. 3 shows a simplified schematic of the gut–brain axis and the appetite and homeostatic brain areas modulated by food intake.

Fig. 3. (colour online) A simplified illustration of the gut–brain axis and appetite and homeostatic brain areas modulated by food intake. OFC, the orbitofrontal cortex; ACC, anterior cingulate cortex; PYY, peptide YY, GLP-1, glucagon-like peptide 1; CCK, chole cystokinin.

Peptide YY (PYY), glucagon-like peptide 1 and cholecystokinin (CCK) are anorectic hormones secreted by the gut and intestine following a meal ingestion, leading to inhibition of food intake. Prior BOLD-fMRI studies have assessed the brain response to an intravenous infusion of PYY, and showed a modulation in OFC, posterior insula, ACC, striatum and homeostatic regions of the brainstem and hypothalamus⁽ Reference Batterham, Ffytche and Rosenthal ^{81 )}. The subjects’ energy intake correlated negatively with OFC signal change after the PYY infusion and positively with hypothalamus activity after saline infusion. These authors postulated the presence of PYY switched the regulation of food intake from a homeostatic brain region (hypothalamus) to a hedonic region (OFC). In a more recent study, the combined administration of PYY and glucagon-like peptide 1 to fasted human subjects attenuated brain areas that control appetitive behaviour, including amygdala, caudate, insula, nucleus accumbens, OFC and putamen⁽ Reference De Silva, Salem and Long ^{82 )}.

CCK is released within minutes of eating, as a function of the presence of fat (long-chain NEFA) or protein, and thought to play a role in meal termination⁽ Reference Moran and Kinzig ^{83 ,} Reference Liddle, Goldfine and Rosen ^{84 )}. The satiety effects of CCK are mediated in part by the vagal nerve⁽ Reference Batterham, Cowley and Small ^{85 ,} Reference Degen, Matzinger and Drewe ^{86 )}, to activate the brainstem nucleus tractus solitarius which projects to brain regions controlling food intake (e.g. hypothalamus and amygdala)⁽ Reference Petrovich, Setlow and Holland ^{87 )}. Very few studies have assessed the brains’ response to nutrients and the role of CCK in mediating such a response. The intra-gastric infusion of lipids has been shown to lead to a CCK-dependent increase in BOLD signal in the brainstem, pons, hypothalamus, cerebellum and motor cortical areas⁽ Reference Lassman, Mckie and Gregory ^{88 )}. Li et al. investigated the brain response to different macronutrients (fat, glucose and protein) and measured the associated gut hormone concentrations, including ghrelin (appetitive hormone), CCK, and glucagon-like peptide 1, in both fed and fasted states⁽ Reference Li, An and Zhang ^{89 )}. The BOLD response reduced when fed in the middle insula, thalamus, parahippocampal cortex, caudate, and lateral OFC, compared with the hunger state⁽ Reference Li, An and Zhang ^{89 )}. However, amygdala activation was modulated following protein ingestion but not after fat or glucose, suggesting it may play a more important role than other regions in mediating dietary protein-induced satiety. BOLD signal changes were positively correlated with circulating ghrelin concentrations and were negatively correlated with circulating insulin, CCK and glucagon-like peptide 1 concentrations. More recently, the CCK plasma values have been showed to correlate negativity with BOLD response in reward (amygdala) and taste (insula) regions in response to ingestion of a high-fat meal⁽ Reference Eldeghaidy, Marciani and Hort ^{90 )}. A number of studies have also investigated the cortical response to ghrelin⁽ Reference Sun, Veldhuizenb and Wray ^{91 –} Reference Malik, McGlone and Bedrossian ^{93 )}.

Effect of obesity on the cortical response

There is now extensive literature on using fMRI to examine how obesity influences the cortical response to food. Recent fMRI studies showed that brain activity of obese people is significantly different from that of normal-weight control subjects in several brain regions implicated in food reward, emotion/memory and sensory/motor processing, with greater activation in the obese group paired with hypo-activity in areas associated with homeostatic satiety and cognitive control/attention. Szalay et al. showed that obese subjects had high levels of activation in taste and reward areas (amygdala, nucleus accumbens for sweet, fatty or quinone samples, whereas control healthy subjects varied their response depending on the taste of the sample⁽ Reference Szalay, Aradi and Schwarcz ^{94 )}. The authors postulated that the dysfunction in reward circuitry in obese subjects may have a distinguished role in overeating due to their altered responsiveness to tastes. Other studies⁽ Reference Rothemund, Preuschhof and Bohner ^{95 ,} Reference Stoeckel, Weller and Cook ^{96 )} demonstrated that food motivation associated with viewing high- and low-energy food images in obese women is different from healthy weighted subjects, with an enhanced activation to high-energy food images in dorsal striatal regions (reward areas). Moreover, increasing body BMI positively correlated with BOLD signal for the high-energy condition in anterior insula (taste region), OFC (reward and secondary taste area), posterior cingulate, dorsal striatum and post-central gyrus. A more recent functional connectivity analysis study showed a relative deficiency in the reward network in obese women, with a deficit in amygdala modulated activation of both OFC and nucleus accumbens, but a tendency towards excessive modulation of the nucleus accumbens by the OFC⁽ Reference Stoeckel, Kim and Weller ^{97 )}. These results show it may not only be the hyper-activation of the reward system in obese individuals that increases the motivational values of food, but also differences in the interaction of regions within the reward network. Together these studies demonstrate unique patterns of brain activation in obese compared to healthy control individuals when food motivation is manipulated by varying food content.

What can ultra-high field offer to functional MRI studies of nutrition?

The studies described in this review demonstrate the application of fMRI to reveal brain regions involved in food perception, and provide a powerful method to study the interactions between homeostatic and hedonic signals controlling eating mechanism. In addition, these studies suggest that the neural circuitry involved in the control of food intake is not limited to the hypothalamus and brainstem, but involves hedonic brain regions. However, there is still much research needed to understand the neural underpinnings in nutrition research in the human brain. One of the major gains of UHF (7 T) MR is the increased BOLD contrast-to-noise ratio. The use of such UHF scanners will improve the spatial mapping, and allow better differentiation of cortical areas involved in uni-modal and multi-modal processing, for example in the primary gustatory cortex. Furthermore, the increased BOLD contrast-to-noise ratio can be applied to study single trial BOLD responses and habituation, thus allowing the delivery of more natural stimuli and reduced adaptation effects.