Results
The search strategy identified 1586 articles, 216 articles were screened for inclusion with 64 articles describing 60 studies included in the final analysis (26–29, 33–92) as described in Table S1 in Supplementary Material. The primary reasons for exclusion were: the article did not meet inclusion criteria for study design (n = 60); no relevant outcome was studied (n = 58); and the study investigated a population not specified in the inclusion criteria (n = 35). Three additional studies were excluded as they did not report BMI or weight status of participants (93–95).
A total of 1565 participants were included across the studies (mean 26, range 5–100). Age ranged from 18–66 years with the most commonly studied age group being 18–35 years olds (n = 42 studies) (26, 28, 29, 33–43, 45–47, 60–66, 68–72, 74–86, 89, 91, 92). Participants were predominantly female and right handed with 26 of the studies exclusively recruiting females (26, 29, 33, 43, 45–48, 56–59, 65, 66, 70–72, 75, 76, 78–81, 83–85, 90–92). The majority of studies were published post 2009 (n = 53, 83%) and used a within participants cross-over design (n = 25, 42%). No randomized control trials were retrieved by the search criteria.
Seventeen of the included studies compared both obese and healthy weight participants in the same study (26–28, 33–47). Twelve reported outcomes from pre- to post-weight loss (48–59). Methods of weight loss included bariatric surgery (n = 7) (48, 49, 52, 53, 57–59) and behavioral nutrition and lifestyle interventions (n = 5) (51, 53–56) with follow up in these studies ranging from one month to twelve months. Weight loss ranged from 3.4–25% of original body mass in these studies. Five of the included studies were exclusively conducted in overweight or obese participants (BMI ≥ 25.00 kg/m2) (88–92) and 26 studied individuals with a mean BMI in the healthy weight range (BMI < 25.00 kg/m2) (29, 60–87). However, eight of the healthy weight studies included participants with BMI’s spanning from the underweight category to the overweight/obese category (29, 67, 69, 74, 78, 79, 81, 82).
As outlined in Table S1 in Supplementary Material, 30 studies (50%) used participants who were fasted prior to fMRI scans (range 2–24 h) (26–29, 40–42, 44–46, 50, 51, 54, 55, 64, 67, 73, 75, 77–82, 86, 87, 90–92). Seven studied satiated participants (43, 47, 48, 58–60, 76, 83–85) and 23 investigated neural responses in both fasted and satiated conditions (27, 33–39, 52, 53, 56, 57, 61–63, 65, 66, 68–72, 74, 88, 89). The most common variables other than fMRI assessed in the included studies were hunger (n = 34 studies), appetite (n = 10 studies), and liking ratings of presented foods (n = 10 studies).
Food images used in the studies were described by their authors as “high-calorie” foods in 36 of the 60 studies (26, 28, 29, 35, 38, 40, 41, 43–50, 54–61, 64–66, 68, 69, 71, 76, 78–81, 83–88, 90–92) and included foods such as chocolate, chips, and hamburgers. Foods described as “low-calorie” foods were used in 32 studies (28, 35, 38, 40, 42, 44, 46–48, 50, 54–61, 64–66, 68, 69, 71, 78–81, 83–88, 90–92) and included foods such as fruit and vegetables. The actual calorific values for foods were only reported in seven of the studies (35, 47, 57–59, 78, 88). No studies reported the use of a dietitian or nutritionist in the selection or classification of foods. Fifteen studies used foods based on the appeal and salience of the food (e.g., “hedonic,” “palatable,” and “appetizing”) (27, 33, 34, 41, 43, 52, 62, 63, 70, 73, 75, 77–79, 81, 82). Food images were selected using pilot ratings of palatability, perceived calorific value, and recognizability of presented images in only 19 studies (27, 36, 37, 40, 43, 44, 52, 53, 61, 68, 70, 71, 74, 77, 79–82, 86). Control images were used in the majority (n = 48) of the included studies and varied greatly, including images of cars, office equipment, landscapes, and blurred images.
Block design was used in 38 of the studies (27, 28, 34, 35, 37–42, 44, 46–53, 55, 56, 60–64, 67–69, 71–73, 81–85, 87–89, 91, 92) and a 3 T magnet was used most commonly to acquire imaging data (n = 43) (26, 28, 29, 34, 36, 37, 39–46, 49–53, 55, 56, 60–63, 65–69, 71–73, 75, 76, 78–82, 87, 89–92) (Table S2 in Supplementary Material). The imaging plane most commonly used to acquire images of the brain was the transverse plane parallel to the anterior commissure posterior commissure line (AC–PC line) (n = 18).The method of reporting fMRI results was variable across the range of studies. All studies excluding two reported Talairach or MNI coordinates, but only 32 (53%) studies reported cluster size or volume of activation (35, 38, 39, 41, 43, 44, 46, 49, 50, 52, 53, 55–59, 62–64, 66–69, 72, 74, 77, 78, 81, 84–86, 88, 91).

Healthy weight compared to overweight/obese participants
Across studies comparing overweight and obese participants to healthy weight controls, overweight/obese individuals had greater brain activity to foods compared to non-foods in areas associated with a variety of functions in the context of food-cue processing (96). This included areas associated with reward processing [insula (26, 33, 41, 43, 47), orbitofrontal cortex (OFC) (26, 28, 43)], reinforcement and adaptive learning [amygdala (27, 28, 33, 43), putamen (28, 41, 47), OFC (26, 28, 43)], emotional processing [insula (26, 33, 41, 43, 47), amygdala (27, 28, 33, 43), cingulate gyrus (44, 45)], recollective, and working memory [amygdala (27, 28, 33, 43), hippocampus (27, 28, 33, 45), thalamus (33, 41), posterior cingulate cortex (27, 47), caudate (28, 45, 47)], executive functioning [prefrontal cortex (PFC) (28, 43), caudate (28, 45, 47), cingulate gyrus (44, 45)], decision making [OFC (26, 28, 43), PFC (28, 43), thalamus (33, 41)], and visual processing [thalamus (33, 41), fusiform gyrus (27, 43)]. Additionally, obese individuals displayed greater activation to food cues in areas involved in motor learning and coordination such as hand-to-mouth movements and swallowing [insula (26, 33, 41, 43, 47), putamen (28, 41, 47), thalamus (33, 41), caudate (28, 45, 47)] as well as risk aversion [inferior frontal gyrus (41, 44)]. These increases in brain activity were particularly evident in response to high-calorie foods compared to low-calorie foods. When satiated, increased activity was observed in obese compared to healthy weight individuals in areas involved in decision making [PFC (34, 39), OFC, and caudate (38)], reward anticipation [anterior cingulate (33, 38) and OFC (33, 38)] as well as emotional processing [insula (33, 37), caudate (38), and amygdala (33)]. Significant correlations between BMI and activation were reported in three studies (37, 44, 47).

Weight change interventions
In three studies using a nutrition and lifestyle intervention, brain activation at the commencement of the intervention was associated with degree of weight-loss success and maintenance. This included areas associated with reward processing and anticipation [insula, anterior cingulate cortex, nucleus accumbens (55), and the OFC (56)], decision making [PFC (54) and OFC (56)], and impulsivity [nucleus accumbens and the anterior cingulate cortex (55)]. Participants who had successfully lost and maintained weight-loss displayed differential neural responses to food cues to those of healthy weight participants in areas involved in emotion, memory, and visual processing [cingulate gyrus, parietal cortex (51)], and to that of obese participants in regions associated with emotion, impulse control, and reward-based learning [PFC and the anterior cingulate (50)].
In studies reporting pre- to post-bariatric surgery outcomes, reductions in activity were reported in the insula and putamen. These areas are implicated in interoceptive processing (52, 58) and reinforcement learning (57, 59), respectively. Further, activation of the hypothalamus, which regulates hunger and subsequent food intake, following gastric bypass surgery resembled the responses of healthy weight individuals more closely than responses of obese individuals. More successful weight loss in gastric bypass surgery was associated with increased baseline neural activity of the dorsolateral PFC (49) and unique changes in activity were found depending on the method of weight loss (i.e., behavioral or surgical) (53). A relationship between BMI and activation of areas involved in reward anticipation and impulsivity [anterior cingulate cortex (49) and middle frontal gyrus (52)] was identified in two studies.

Healthy weight participants
The most common finding across studies of healthy weight participants was that motivational state (i.e., fasted or satiated state) affected brain activation to food. Fasting often increased responses to high-calorie foods in areas associated with processing of reward and stimuli salience [OFC (66, 69, 71, 72), striatum (65, 69, 72), insula (69, 71)], decision making [OFC (66, 69, 71, 72), striatum (65, 69, 72)], implicit learning [OFC (66, 69, 71, 72), putamen (66, 71)], and the processing of visual cues [fusiform gyrus (68, 71, 74)]. Gender differences were identified in responses to food cues with females displaying greater activation in a variety of brain regions implicated in attention, emotion, recollective memory, and decision making (67, 68, 74).

Obese participants
In obese participants, food compared to non-food images activated areas including the PFC, insula, amygdala, nucleus accumbens (91, 92), and cerebellum (89). These areas are associated with numerous roles which could affect food cue processing including executive functioning, reward processing, and anticipation, reinforcement learning, memory modulation, and motor control. Females showed greater activation in the caudate and OFC when fasted and greater activation in the anterior cingulate cortex when satiated (88). Abdominal adiposity predicted brain activity in one study (92).

Results of the ALE meta-analysis
As only one study reported increases in brain activation following weight loss, only studies reporting decreases in neural activation from pre- to post-weight loss were included in the meta-analysis. Five studies describing seven experiments were identified that met the meta-analysis inclusion criteria with 45 participants and 41 foci (55–59). The meta-analysis identified 13 clusters, which survived statistical thresholds, as demonstrated in Figure 2. The largest cluster was the left superior temporal gyrus (MNI: −40, −48, 6), as described in Table S3 in Supplementary Material. Other clusters surviving statistical thresholds included right middle frontal gyrus (MNI: 32, 34, 34), left lentiform nucleus (MNI: −12, 0, −2), left cingulate gyrus (MNI: −4, −34, 26,), and right precentral gyrus (MNI: 40, 0, 42).
Figure 2  Axial (z), coronal (y), and sagittal (x) views of decreased activation in studies comparing neural activation to visual food cues from pre- to post-weight loss, as detected by fMRI. FDR corrected P  < 0.05, cluster size >100 mm3, region of interest studies excluded. Figure shows decreased activation from pre- to post-weight loss in the cingulate gyrus, middle frontal gyrus, and precuneus.