Wavelet-enhanced ICA

A hybrid that removes artifact energy from a component instead of deleting the whole component. After ICA, artifactual components are wavelet-thresholded: the large, artifact-related wavelet coefficients are suppressed while small coefficients — assumed to carry neural activity that leaked into the same component — are kept, and the denoised component is back-projected. This addresses ICA's blunt all-or-nothing removal, which discards any brain signal captured by a component judged artifactual.

wICA and related partial-removal schemes are useful when artifacts and brain sources are imperfectly separated (few channels, short recordings) so that no component is purely artifact. The tradeoff is added assumptions and parameters (wavelet basis, threshold rule) and the risk of leaving residual artifact if thresholds are lax; it is one of several correction-not-rejection tools alongside ASR and SSP.