Test-Retest Reproducibility of the Intrinsic Default Mode Network: Influence of Functional Magnetic Resonance Imaging Slice-Order Acquisition and Head-Motion Correction Methods.

Head motion is a known challenge in resting-state functional magnetic resonance imaging studies for biasing functional connectivity (FC) among distinct anatomical regions. These persist even with small motion, limiting comparisons of groups with different head-motion characteristics. This motivates an interest in the optimization of acquisition and correction strategies to minimize motion sensitivity. In this test-retest (TRT) study of healthy young volunteers (N = 23), we investigate the effects of slice-order acquisitions (sequential or interleaved) and head-motion correction methods (volume- or slice-based) on the TRT reproducibility of intrinsic connectivity of the default mode network (DMN). We evaluated the TRT reproducibility of the entire DMN and each main node using the absolute percentage error, intraclass correlation coefficient (ICC), and the Jaccard coefficient. Regardless of slice-order acquisition, the slice-based motion correction method systematically estimated larger motion and returned significantly higher temporal signal-to-noise ratio. Although consistently extracted across all acquisition and motion correction approaches, DMN connectivity was sensitive to these choices. However, the TRT reproducibility of the whole DMN was stable and showed no sensitivity to the methods tested (absolute reproducibility ∼7%, ICC = 0.47, and Jaccard = 40%). Percentage errors and ICCs were consistent across single nodes, but the Jaccard coefficients were not. The posterior cingulate was the most reproducible node (Jaccard = 52%), whereas the anterior cingulate was the least reproducible (Jaccard = 30%). Our study suggests that the slice-order and motion correction methods evaluated offer comparable sensitivity to detect DMN connectivity changes in a longitudinal study of individuals with low head-motion characteristics, but that controlling for the consistency in acquisition and correction protocols is important in cross-sectional studies.