Equitable Thresholding and Clustering

TL;DR: The Equitable Thresholding and clustering (ETAC) as mentioned in this paper approach is based on two ideas: (i) reducing the dependence of clustering results on arbitrary parameter values by using multiple sub-tests, and (ii) adjusting the cluster-thresholding parameters of each sub-test in an equitable way.

Abstract: We describe a novel hybrid method to threshold FMRI group statistical maps derived from voxelwise second-level statistical analyses. The proposed Equitable Thresholding and Clustering (ETAC) approach is grounded in two ideas: (i) reducing the dependence of clustering results on arbitrary parameter values by using multiple sub-tests -- each equivalent to a "standard" FMRI clustering analysis -- to make decisions about which groups of voxels are potentially "significant" then combining the results of each sub-test to decide which voxels are "accepted"; and (ii) adjusting the cluster-thresholding parameters of each sub-test from (i) in an equitable way -- so that the individual false positive rates (FPRs) are balanced across sub-tests and voxels -- to achieve a desired global FPR (e.g., 5%). ETAC is independent of parametric assumptions about the spatial correlation of FMRI noise, because resampling methods are used to simulate the null (noise-only) distribution required to compute the FPR evaluations required in (ii). Resting FMRI datasets, analyzed with pseudo-task timings to provide a null model, were used to show the accuracy of the ETAC FPR control. A task FMRI data collection was used to compare ETAC's true positive detection power vs. a standard cluster detection method, with ETAC providing equivalent or favorable results. Additionally, an important general note on the use of one-sample t-tests in neuroimaging is also made, based on an examination of the variability of cluster results.