Schedule of Events | Symposia

BrainEffeX: A web app for exploring fMRI effect sizes

Poster Session D - Monday, March 31, 2025, 8:00 – 10:00 am EDT, Back Bay Ballroom/Republic Ballroom
Also presenting in Data Blitz Session 4 - Saturday, March 29, 2025, 10:30 am – 12:00 pm EDT, Constitution B.

Hallee Shearer¹ (h.shearer@northeastern.edu), Matt Rosenblatt², Jean Ye², Rongtao Jiang², Link Tejavibulya², Qinghao Liang², Javid Dadashkarimi^3,6, Margaret Westwater², Iris Cheng², Alexandra Fischbach¹, Ashley Humphries⁴, Aneesh Kumar¹, Max Rolison², Hannah Peterson², Brendan Adkinson², Saloni Mehta², Chris Camp², Thomas Nichols⁵, Joshua Curtiss¹, Dustin Scheinost², Stephanie Noble¹; ¹Northeastern University, ²Yale University, ³Massachusetts General Hospital, ⁴University of Nebraska-Lincoln, ⁵University of Oxford, ⁶Harvard Medical School

Estimating effect size is a critical step in power analyses, and can help inform experimental design. However, effect size estimation is particularly difficult for fMRI data due to the complexity of both the data and the analysis techniques. Further, it is difficult to obtain estimates from the literature, and small sample sizes of pilot studies may not provide precise enough estimates. When similar studies can be found in the literature, effect sizes are often not reported across the whole brain, limiting utility for study design. To facilitate the estimation and exploration of effect sizes for fMRI, we estimated effects for “typical” study designs with large (n>500) datasets (ABCD, HCP, HBN, PNC, UKB). We conducted brain-behavior correlations, task vs. rest contrasts, and between-group analyses with both functional connectivity and task-based activation maps. The analyses leverage fMRI data from rest and commonly used tasks, and behavioral data reflecting various phenotypes. In light of recent research supporting the promise of broader-level methods, we included network-level and multivariate versions of all analyses. We repeated analyses with four motion deconfounding strategies: statistical control, full residualization, thresholding, and no correction. We transformed results to Cohen’s d and R-squared estimates of effect size and calculated simultaneous confidence intervals. Finally, we created an interactive web application (BrainEffeX) for comprehensively exploring and visualizing these results. BrainEffeX is the first step in an effort to address the need for facilitated power calculations in fMRI by providing a growing resource enabling researchers to estimate and summarize effect sizes for fMRI studies.

Topic Area: METHODS: Neuroimaging