CMStatistics 2021: Start Registration
View Submission - CMStatistics
Title: Permutation testing of network enrichment in neuroimaging Authors:  Sarah Weinstein - University of Pennsylvania (United States) [presenting]
Aaron Alexander-Bloch - University of Pennsylvania (United States)
Simon Vandekar - Vanderbilt University (United States)
Erica Baller - University of Pennsylvania (United States)
Ruben Gur - University of Pennsylvania (United States)
Raquel Gur - University of Pennsylvania (United States)
Armin Raznahan - National Institute of Mental Health Intramural Research Program (United States)
Azeez Adebimpe - University of Pennsylvania (United States)
Theodore Satterthwaite - University of Pennsylvania (United States)
Russell Shinohara - University of Pennsylvania (United States)
Abstract: In studies of neurodevelopment, anatomical and functional brain parcellations are often used to analyze patterns of association between a phenotype (e.g., age) and an imaging feature (e.g., cortical thickness) and determine whether those associations are especially strong or ``enriched'' in those subregions. However, existing methods in this area may not be reliable, as they do not account for the spatial structure of the data, leading to inflated type I errors. We propose Network Enrichment Analysis Testing (NEAT), which adapts Gene Set Enrichment Analysis (GSEA), widely used in genomics research, to statistically test whether associations between high-dimensional imaging features and non-imaging phenotypes are enriched within functional networks or other parcellations of the brain. NEAT incorporates random permutations of subject-level data and augments imaging measurements with spatial smoothing to enhance statistical power in the assessment of network-specific enrichment, while preserving conservative type I error rates. We illustrate the properties of NEAT in simulated and real neuroimaging data from the Philadelphia Neurodevelopmental Cohort.