Regressor Selection for RETROICOR and RETROKCOR Corrections of 3D FMRI Data

Rob H.N. Tijssen¹, ², Mark Jenkinson¹, Karla L. Miller¹

In this work we explore the use of retrospective corrections, in order to remove temporal instabilities from 3D FMRI acquisitions. Using the Bayesian Information Criterion (BIC) to determine an optimal subset, previously proposed for RETROICOR, we aim to optimize the regressor set for various 3D acquisition methods (e.g., SPGR and bSSFP). The results suggest that the optimal number of regressors is highly dependent on the acquisition and can even be zero (i.e., no retrospective correction) in some cases (e.g., when a real-time cardiac synchronized readout is used to correct for cardiac fluctuations prospectively).