Adaptive Noise Removal IRF-RETROICOR

Erik B. Beall¹, Mark J. Lowe¹

RETROICOR is a good general model for physiologic noise, however it is inefficient and provides no information about actual temporal signatures of noise. We show modifications that provide these signatures and that a small subset of these account for all the RETROICOR modeled noise variance without removing as much non-noise (signal of interest) variance. We show remarkable concurrence of these 4 cardiac and 2 respiratory signatures across 34 subjects, in both fMRI and resting connectivity data. Based on the stability of these, we propose that they may be used to study abnormal physiology of cardiac and respiratory coupling.