Reducing Effects of Drift in fMRI Data Using Joint Reconstruction of R2* and Field Maps

Introduction In blood oxygenation level dependent (BOLD) functional MRI (fMRI), linear signal drifts are often seen in the time series. These drifts are nuisance parameters that are either regressed or filtered out when performing the functional analysis. The largest sources are believed to be scanner instability such as B0 drift and physiological noise and motion [1]. B0 drift also causes drift in off-resonance that can be measured by estimating field maps for all time frames in fMRI data. The field maps can then be used in off-resonance corrected reconstructions to correct for these temporal changes. We recently proposed a method for jointly reconstructing R2* and field maps from single shot k-space data [2]. These field drift corrected R2* estimates should be less sensitive to B0 drifts. Here we investigate the reduction in drift in R2* maps from this reconstruction and compare it to results from conventional BOLD-weighted fMRI analysis. Joint Reconstruction of Dynamic R2* and field maps The joint reconstruction [2] is based on estimating R2* and field map for each time frame by repeated refinements using linear approximations. For an fMRI study, time frame j of acquired k-space data j y can be modeled as follows: