HINT: A Toolbox for Hierarchical Modeling of Neuroimaging Data
share
The modular behavior of the human brain is commonly investigated using independent component analysis (ICA) to identify spatially or temporally distinct functional networks. Investigators are commonly interested not only in the networks themselves, but in how the networks differ in the presence of clinical or demographic covariates. To date, group ICA methods do not directly incorporate these effects during the ICA decomposition. Instead, two-stage approaches are used to attempt to identify covariate effects (Calhoun, Adali, Pearlson, and Pekar, 2001; Beckmann, Mackay, Filippini, and Smith, 2009). Recently, Shi and Guo (2016) proposed a novel hierarchical covariate-adjusted ICA (hc-ICA) approach, which directly incorporates covariate information in the ICA decomposition, providing a statistical framework for estimating covariate effects and testing them for significance. In this work we introduce the Hierarchical Independent Component Analysis Toolbox, HINT, to implement hc-ICA and other hierarchical ICA techniques.
READ FULL TEXT