Even at rest, functional magnetic resonance imaging (fMRI) data displays exquisitely complex temporal dynamics. Here, we deployed a time-frequency analysis to track the modulus of fMRI signals over time, across space (a set of 341 brain areas) and frequency (45 uniformly distributed bins in the 0.01-0.25 Hz range).

Decomposing the data into a set of temporally overlapping building blocks by Principal Component Analysis, we exposed diverse functional components with their own modulus pattern across brain locations and frequency sub-ranges.

In particular, the component explaining most data variance showed homogeneous modulus across space at low frequencies, fitting with the marked whole-brain signal fluctuations seen in the time courses subjected to analysis. Other components showed topologically well-defined modulus patterns (e.g., contrasting the default mode and visual networks), with characteristic frequency properties and subject-specific activation profiles.