Connectome caricatures: large-amplitude co-activation patterns in resting-state fMRI hide sources of individual differences

Motivation: Task-like co-activation patterns are sparsely present during resting-state fMRI but drive functional connectivity. However, little research has characterized the remaining signal. Goal(s): We aimed to characterize the hidden resting-state signal that exists beyond the dominating co-activation patterns and assess its merit for studying individual differences. Approach: We used task-based fMRI data to construct a task-relevant co-activation pattern manifold. By projecting resting-state time series data orthogonally to this manifold, we created Caricatured connectomes. Results: Like caricatures, these connectomes emphasized individual differences while reducing between-individual similarity. They also represented individual differences in behavior, often to a greater degree than Standard connectomes. Impact: A distinct signal carrying information about individual differences exists beyond the dominating co-activations that drive resting-state functional connectivity. This signal may better characterize the brain's intrinsic functional architecture and can be used to evaluate novel sources of individual differences.