Robust R ∗ 2 map estimation from motion scattered slices for fetal fMRI

A major challenge for fMRI analysis of the developing brain is subject motion, which can corrupt T 2 ∗ -weighted (T 2 ∗ -w) signal intensity with spin history effects. Multi echo multislice EPI acqusitions can be used to create parameteric R 2 ∗ mapping for fMRI that can provide independence from such signal variation. However, motion between slice acquisitions scatters the measurements over the anatomy of interest at each time frame. Such motion scattering cannot be optimally accounted for using direct interpolation, as data at each voxel can be missing or duplicated due to motion. Conventional BOLD contrast techniques have recently employed iterative time series reconstruction schemes to make best use of acquired data. Here we propose an extension to such techniques to deal with the case of multi-echo parametric fitting from scattered slices. Specifically, we formulate a robust R ∗ 2 map estimation scheme for scattered multi-slice slice data, making use of robust fitting of R 2 ∗ values to the spatially scattered T 2 ∗ -w slices. We present results on simulated data and real in-utero imaging that indicate the approach outperforms the conventional interpolation based recovery of R ∗ 2 timeseries.