tensor-decomposition-brain-states

You can identify quasi-stationary brain states from dynamic connectivity data by applying tensor decomposition to a time-resolved connectivity tensor, extracting temporally invariant network states and their topographic representations.

Yes, tensor decomposition for dynamic brain connectivity is cross-modal and applies to EEG, fMRI, and MEG datasets to uncover quasi-stationary network states during cognitive tasks or resting-state analysis.

Time-varying connectivity data must be constructed as a connectivity tensor with dimensions Time × Region × Region before applying Tucker or CP decomposition to extract brain network states.

Both Tucker and CP decomposition extract temporally invariant network states from dynamic connectivity tensors, with Tucker allowing flexible component counts per mode and CP yielding a direct sum of rank-one factors for state mapping.

After tensor decomposition, you can visualize brain network states using optional state mapping and topographic visualization features, generating topographic maps per state for interpretability of the extracted quasi-stationary connectivity patterns.

Use tensor decomposition for dynamic brain connectivity when you need a compact representation of quasi-stationary network states from time-varying data, enabling state discovery and transition analysis that scalar methods cannot capture.