ABSTRACT: Anorexia nervosa (AN) is a severe mental disorder characterized by food restriction and weight loss. This study aimed to test the model posed by Brooks et al. (2012a,b) that women suffering from chronic AN show decreased food-cue processing activity in brain regions associated with energy balance and food reward (bottom-up; BU) and increased activity in brain regions associated with cognitive control (top-down; TD) when compared with long-term recovered AN (REC) and healthy controls (HC). Three groups of women, 15 AN (mean illness duration 7.8?±?4.1 years), 14 REC (mean duration of recovery 4.7?±?2.7?years) and 15 HC viewed alternating blocks of food and non-food images preceded by a short instruction during functional magnetic resonance imaging (fMRI), after fasting overnight. Functional region of interests (fROIs) were defined in BU (e.g., striatum, hippocampus, amygdala, hypothalamus, and cerebellum), TD (e.g., medial and lateral prefrontal cortex, and anterior cingulate), the insula, and visual processing areas (VPA). Food-cue processing activation was extracted from all fROIs and compared between the groups. In addition, functional connectivity between the fROIs was examined by modular partitioning of the correlation matrix of all fROIs. We could not confirm the hypothesis that BU areas are activated to a lesser extent in AN upon visual processing of food images. Among the BU areas the caudate showed higher activation in both patient groups compared to HC. In accordance with Brooks et al.'s model, we did find evidence for increased TD control in AN and REC. The functional connectivity analysis yielded two clusters in HC and REC, but three clusters in AN. In HC, fROIs across BU, TD, and VPA areas clustered; in AN, one cluster span across BU, TD, and insula; one across BU, TD, and VPA areas; and one was confined to the VPA network. In REC, BU, TD, and VPA or VPA and insula clustered. In conclusion, despite weight recovery, neural processing of food cues is also altered in recovered AN patients.
