Impacts of Simultaneous Multislice Acquisition on Sensitivity and Specificity in fMRI

Abstract

Simultaneous multislice (SMS) acquisition can be used to decrease the time between acquisition of fMRI volumes, which can increase sensitivity by facilitating the removal of higherfrequency artifacts and boosting effective sample size. The technique requires an additional processing step in which the slices are separated, or unaliased, to recover the whole brain volume. However, this may result in signal "leakage" between aliased locations, i.e., slice "leakage," and lead to spurious activation (decreased specificity). SMS can also lead to noise amplification, which can reduce the benefits of decreased repetition time. In this study, we evaluate the original slice-GRAPPA (no leak block) reconstruction algorithm and acquisition scenarios used in the young adult Human Connectome Project (HCP), as well as split slice-GRAPPA (leak block). In addition to slice leakage, signal leakage can result from spatial smoothing, i.e., smoothing leakage, which leads to inflated regions of activation. Previous studies have generally found that SMS acquisition results in higher test statistics and/or a greater number of activated voxels. Here, we use simulations to disentangle this phenomenon into true positives (sensitivity) and false positives (decreased specificity). Slice leakage was greatly decreased by split slice-GRAPPA. Noise amplification was decreased by using moderate acceleration factors (AF = 4). We examined slice leakage in unprocessed fMRI motor task data from the HCP, which used the original slice-GRAPPA. When data were smoothed, we found evidence of slice leakage in some, but not all, subjects. We also found evidence of SMS noise amplification in unprocessed task and processed resting-state HCP data.

Cluster failure revisited: Impact of first level design and physiological noise on cluster false positive rates

“…We note that multiband MR sequences (Moeller et al, ) are becoming increasingly common to improve temporal and/or spatial resolution, for example as provided by the Human Connectome Project (Essen et al, ) and the enhanced NKI‐Rockland sample (Nooner et al, ). Multiband data have a potentially complex spatial autocorrelation (see, e.g, Risk, Kociuba, and Rowe ()), and an important topic for future work is establishing how this impacts parametric cluster inference. The nonparametric permutation test (Winkler et al, ) does not make any assumption regarding the shape of the SACF, and is therefore expected to perform well for any MR sequence.…”

Accurate modeling of temporal correlations in rapidly sampled fMRI time series

“…As the TR was reduced with increasing MB factor, the flip angle was optimized to match the Ernst angle based on a grey matter (GM) T1 value of 1,000 ms at 3 T (Weiskopf et al, 2013). All multiband RF excitations were performed with MB RF Phase Scramble selected (Wong 2012) and the data were reconstructed using the MB LeakBlock Kernel option (Cauley et al, 2014), which has been shown to suppress residual aliasing of BOLD signal across slices in fMRI (Risk, Kociuba, & Rowe, 2018;Todd et al, 2016). The same echo-time was chosen for all the protocols (TE 5 30.2 ms).…”

Temporal signal-to-noise changes in combined multiband- and slice-accelerated echo-planar imaging with a 20- and 64-channel coil

“…No spatial smoothing was performed [33]. For group-level statistics, we additionally co-registered each participants' data to the Montreal Neurological Institute (MNI) 2mm standard space and used Freesurfer for surface reconstruction and to visualize statistical maps on inflated surface brains.…”