BrainPowerX - A New Empirical Algorithm for Power Calculation for fMRI

Poster Session F - Tuesday, April 1, 2025, 8:00 – 10:00 am EDT, Back Bay Ballroom/Republic Ballroom

Fabricio Cravo¹, Stephanie Noble^1,2; ¹Northeastern University, ²Yale University

Although fMRI research has transformed our understanding of the human brain, recent studies highlight a critical lack of statistical power (probability of detecting true effects) throughout the field. The high probability of missing true effects limits our ability to learn more about how the brain is functionally organized and poses a major challenge to reproducing results. A proper power analysis enables researchers to mitigate this through sample size estimation and optimization of other elements of analysis for the detection of effects of interest. Existing power calculators rely primarily on parametric methods, which assume prior knowledge about the nature of the underlying probability distribution of the effect. Furthermore, they often exclude more complex inferential procedures often used in neuroimaging, limiting their applicability. To address these limitations, we propose BrainPowerX, a web-based, non-parametric power calculator tailored to typical fMRI studies. It will use data from large, publicly available datasets to estimate true effects and simulate experimental conditions through repeated subsampling. Power will be calculated by measuring the proportion of repetitions where effects are successfully detected in comparison with "ground truth" effects identified from the whole dataset. Preliminary results provide initial power estimates for various inferential methods. Current efforts focus on expanding the flexibility of a Matlab toolbox to accommodate power estimates from more diverse experimental designs. Ultimately, we aim for this toolbox to be an easily-accessible and user-friendly web app that empowers researchers to assess whether their experimental parameters provide sufficient sensitivity to detect their desired effects and adjust experiments accordingly.

Topic Area: METHODS: Neuroimaging