A comprehensive evaluation of multiband-accelerated sequences and their effects on statistical outcome measures in fMRI.

Accelerated functional Magnetic Resonance Imaging (fMRI) with multiband sequences is now relatively widespread. These sequences can be used to dramatically reduce the repetition time (TR) and produce a time-series sampled at a higher temporal resolution. We tested the effects of higher temporal resolutions for fMRI on statistical outcome measures in a comprehensive manner on two different MRI scanner platforms. Experiment 1 tested a range of acceleration factors (1-6) against a standard EPI sequence on a single composite task that maps a number of basic sensory, motor, and cognitive networks. Experiment 2 compared the standard sequence with acceleration factors of 2 and 3 on both resting-state and two task paradigms (an N-back task, and faces/places task), with a number of different analysis approaches. Results from experiment 1 showed modest but relatively inconsistent effects of the higher sampling rate on statistical outcome measures. Experiment 2 showed strong benefits of the multiband sequences on results derived from resting-state data, but more varied effects on results from the task paradigms. Notably, the multiband sequences were superior when Multi-Voxel Pattern Analysis was used to interrogate the faces/places data, but showed less benefit in conventional General Linear Model analyses of the same data. In general, ROI-derived measures of statistical effects benefitted relatively little from higher sampling resolution, with decrements even seen in one task (N-back). Across both experiments, results from the two scanner platforms were broadly comparable. The statistical benefits of high temporal resolution fMRI with multiband sequences may therefore depend on a number of factors, including the nature of the investigation (resting-state vs. task-based), the experimental design, the particular statistical outcome measure, and the type of analysis used. Higher sampling rates in fMRI are not a panacea, and it is recommended that researchers use multiband acquisition sequences conservatively.