Subgroups of eating behavior traits independent of obesity defined using functional connectivity and feature representation learning

AO_SCPLOWBSTRACTC_SCPLOWEating behavior is highly heterogeneous across individuals, and thus, it cannot be fully explained using only the degree of obesity. We utilized unsupervised machine learning and functional connectivity measures to explore the heterogeneity of eating behaviors. This study was conducted on 424 healthy adults. We generated low-dimensional representations of functional connectivity defined using the resting-state functional magnetic resonance imaging, and calculated latent features using the feature representation capabilities of an autoencoder by nonlinearly compressing the functional connectivity information. The clustering approaches applied to latent features identified three distinct subgroups. The subgroups exhibited different disinhibition and hunger traits; however, their body mass indices were comparable. The model interpretation technique of integrated gradients revealed that these distinctions were associated with the functional reorganization in higher-order associations and limbic networks and reward-related subcortical structures. The cognitive decoding analysis revealed that these systems are associated with reward- and emotion-related systems. We replicated our findings using an independent dataset, thereby suggesting generalizability. Our findings provide insights into the macroscopic brain organization of eating behavior-related subgroups independent of obesity.