Less is more: balancing noise reduction and data retention in fMRI with projection scrubbing

Functional MRI (fMRI) data may be contaminated by artifacts arising from a myriad of sources, including subject head motion, respiration, heartbeat, scanner drift, and thermal noise. These artifacts cause deviations from common distributional assumptions, introduce spatial and temporal outliers, and reduce the signal-to-noise ratio of the data--all of which can have negative consequences for the accuracy and power of downstream statistical analysis. Scrubbing is a technique for excluding fMRI volumes thought to be contaminated by artifacts and generally comes in two flavors. Motion scrubbing, based on subject head motion-derived measures, is popular but suffers from a number of drawbacks, among them the need to choose a threshold, a lack of generalizability to multi-band acquisitions, and high rates of censoring of individual volumes and entire subjects. Alternatively, data-driven scrubbing methods such as DVARS are based on observed noise in the processed fMRI timeseries and may avoid many of these issues. In this work, we undertake a comprehensive comparison of motion-based and data-driven scrubbing methods. Because a standardized scrubbing pipeline should work for arbitrary downstream analyses, we argue that the appropriate metric for success is maximal data retention subject to reasonable performance on typical benchmarks such as the reliability and predictiveness of functional connectivity (FC). We also present "projection scrubbing", a novel data-driven scrubbing method based on a statistical outlier detection framework. Projection scrubbing is substantially more beneficial to overall FC reliability than motion scrubbing and DVARS and removes about half the number of volumes, illustrating the advantages of this data-driven, outlier detection approach for identifying contaminated volumes.