Dynamic Missing-Data Completion Reduces Leakage of Motion Artifact Caused by Temporal Filtering That Remains after Scrubbing

Functional magnetic resonance imaging (fMRI) is commonly used to better understand brain function. Data becomes contaminated with motion artifact when a subject moves during an fMRI acquisition. Numerous methods have been suggested to target motion artifacts in fMRI. One of these methods, ?scrubbing?, removes motion-corrupted volumes but must be performed after temporal filtering since it creates temporal discontinuities. Thus, it does not prevent the spread of corrupted time samples from high motion volumes to their neighbors during temporal filtering. To mitigate this spread, which we refer to as ``leakage'', we propose a novel method, Dynamic Missing-data Completion (DMC), that replaces motion-corrupted volumes with synthetic data before temporal filtering. We analyzed the effect of DMC on an exemplary timeseries from a resting state fMRI (rsfMRI) and compared functional connectivity results of six rsfMRI scans from a single subject with different levels of subject motion. Our results suggest that DMC provides added benefit in further reduction of motion contamination that remains after scrubbing. DMC reduced the standard deviation of signal near scrubbed volumes by about 10% compared to scrubbing only, yielding this average closer to that of uncorrupted no motion volumes.