Previous neuroimaging studies explored the structural and functional brain differences in people with obesity, but the results were not fully generalizable due to the scarcity of large-scale, multi-site data. Here, we aimed to construct a neurologic signature of obesity (NSO) using functional magnetic resonance imaging (fMRI) data from large-scale data over 2,500 participants spanning six independent sites.
The resting-state fMRI (rs-fMRI) data of a total of 2,508 participants from six different cohorts were used in this study. 1,497 were obtained from the UK Biobank (UKB) database, 587 were obtained from the Human Connectome Project (HCP) database, 276 were obtained from the enhanced Nathan Kline Institute-Rockland Sample (eNKI-RS) database, 78 were obtained from the St. Vincent’s Hospital (SVH) and Samsung Medical Center (SMC), and 70 were obtained from the Autism Brain Imaging Data Exchange I (ABIDE I) and ABIDE II. Functional connectivity analysis combined with machine learning algorithms was adopted.
The brain regions involved in cognition and reward circuits positively contributed to explaining the characteristics of obesity, while the sensorimotor and visual perception regions showed negative contributions. The generalizability of our NSO model was tested with transferring the model to the totally independent five datasets with different age ranges and ethnicities. The results indicated that the brain regions with strong magnitude largely contribute to explaining obesity (all p < 0.05, FDR corrected).
The NSO model developed in this study may potentially serve as a novel neuroimaging biomarker for obesity.