Study: fMRI Test-Retest Reliability of SST

Test-retest reliability of neural correlates of response inhibition and error monitoring: an fMRI study of a Stop-Signal Task


The present study addressed the following questions: How reliable are fMRI-measured individual differences in brain activation related to response inhibition and error monitoring? Are these brain activations influenced by familial factors? Do regions with stronger group-level task-related activation show higher test-retest reliability (TRR) and familiality (as indicated by MZ twin correlations)? Does longer scanning duration increase TRR? Does within-session reliability relate to longer-term (between-session) reliability? How does in-scanner motion affect the within-subject stability of estimated activation? Does ICA-FIX artifact removal improve the TRR of brain activations?
Fifty-six young adults (32 females, age range: 21-24 years, mean=23.3, SD=0.86) participated in the study. Participants were monozygotic (MZ) twins. All 56 participants (including 27 MZ twin pairs) in the present study completed the first MRI scanning session (Time1), and 44 of them (26 females, 19 MZ twin pairs, age range: 21-24 years, mean=23.3, SD=0.89 at Time1, age range 22-25 years, mean=23.96, SD=0.9 at Time 2) completed a second session approximately 6 months later (Time2, mean (SD) interval 7.9 months (1.70), ranging from 5.7 to 12.0 months). We used a scanner version of the SST developed by (Logan, 1994) and identical to that administered in the ABCD study (Casey et al., 2018). We report reliabilities for the entire brain at the parcel level (MMP1.0, Glasser et al., 2016). Reliability was quantified as the degree of consistency between the Time1 and Time2 measurements, under the assumption of a two-way mixed model, which is known as ICC(3,1) (Shrout and Fleiss, 1979), or alternatively ICC(C,1) (McGraw and Wong, 1996). Intrapair twin correlations were calculated using the same ICC(C,1) estimator as in the TRR estimation.

Response inhibition (RI) and error monitoring (EM) are important processes of adaptive goal-directed behavior, and neural correlates of these processes are being increasingly used as transdiagnostic biomarkers of risk for a range of neuropsychiatric disorders. Potential utility of these purported biomarkers relies on the assumption that individual differences in brain activation are reproducible over time; however, available data on test-retest reliability (TRR) of task-fMRI are very mixed. This study examined TRR of RI and EM-related activations using a the sStop sSignal tTask of the ABCD Study in young adults (n=56, including 27 pairs of monozygotic (MZ) twins) in order to identify brain regions with high TRR and familial influences (as indicated by MZ twin correlations) and to examine factors potentially affecting reliability. We identified brain regions with good TRR of activations related to RI (inferior/middle frontal, superior parietal, and precentral gyri) and EM (insula, medial superior frontal and dorsolateral prefrontal cortex). No subcortical regions showed significant TRR. Regions with higher group-level activation showed higher TRR; increasing task duration improved TRR; within-session reliability was weakly related to the long-term TRR; motion negatively affected TRR, but this effect was abolished after the application of ICA-FIX, a data-driven noise removal method.

Frontiers in Neuroscience - DOI: 10.3389/fnins.2021.624911

  • Ozlem Korucuoglu
  • Michael P Harms
  • Serguei V Astafiev
  • Semyon Golosheykin
  • James T Kennedy
  • Deanna M Barch
  • Andrey P Anokhin
  • Washington University in St. Louis