Simple and efficient steps to improve task-fMRI reliability in the ABCD Study
Poster No:
1068
Submission Type:
Abstract Submission
Authors:
Bader Chaarani1
Institutions:
1UVM, Essex Junction, VT
First Author:
Introduction:
Task-based functional magnetic resonance imaging (fMRI) is a critical neuroimaging technique for understanding brain function. Low reliability is a common concern in task-fMRI analyses because it directly affects the validity and robustness of the findings, including in large datasets such as the ABCD study. The aim of this study is to 1) assess the within-session reliability of task-fMRI data from the ABCD study at each of the four timepoints (ages 9-10; 11-12; 13-14 & 15-16) using intraclass correlation coefficient (ICC) analyses, and 2) test if simple preprocessing steps can improve data reliability.
Methods:
Task-fMRI data from the ABCD study included the primary contrasts from three different cognitive tasks: the Stop Signal task (SST), the N-back task and the monetary incentive delay (MID) task. A rigorous quality control was conducted according to standard ABCD imaging procedures. In addition, participants with poor performance on the tasks were excluded. Vertexwise BOLD activation maps were calculated for each contrast at each time point as the Cohen's D. ICCs were computed at the vertex level, assessing within-session reliability between run 1 and run 2 beta-weights for a given vertex across all participants for each timepoint. Further, ICCs were recalculated after normalizing beta-weights across runs, restricting the vertices to only "active" ones with higher signal (as calculated by Cohen's D) and restricting scans to those with the lowest motion (as evaluated by framewise displacement).
Results:
After applying all quality control steps, over 20,000 scans (~8000 unique children) were included in the longitudinal analyses across the four time points. Overall, all tasks showed robust bilateral activation patterns across timepoints, with the highest activation levels in areas of the cortex known to play a key role in cognitive function (Figure 1). ICCs significantly increased by 40 to 80% across tasks after applying the preprocessing steps, reaching moderate to high reliability levels (Figure 2-A). Moreover, ICC measures of the three tasks improved with age for the majority of contrasts, coupled with improvements in task performance (Figure 2-B). Notably, the N-back contrasts demonstrated the highest reliability among the three tasks at all time points, followed by the SST contrasts. The complete preprocessing code is freely available for the research community.
·Figure 1. Activation maps for each task across four time points, thresholded at Cohen's D >.2.
·Figure 2. A- Mean ICC values after each processing step in the Nback task. B) Mean ICC for all tasks/contrasts across time after restricting the analyses to only include active areas (Cohen's D>.2).
Conclusions:
These findings suggest that task-fMRI reliability improves after simple yet efficient data preprocessing steps in the ABCD dataset. Researchers may consider preprocessing their data and restricting their analyses to tasks/contrasts and regions, vertices, or voxels with higher reliability to enhance the statistical robustness of neuroimaging research.
Higher Cognitive Functions:
Executive Function, Cognitive Control and Decision Making 2
Learning and Memory:
Working Memory
Modeling and Analysis Methods:
Activation (eg. BOLD task-fMRI) 1
Methods Development
Keywords:
Development
FUNCTIONAL MRI
1|2Indicates the priority used for review