Advocating for well-defined and validated procedures: Comment on Griffanti et al., Neuroimage 154:188-205.

Overview

Griffanti et al. (2017) provide a rich set of expert-consensus guidelines for how to label components derived from ICA as either belonging to signal of interest or noise to be filtered from fMRI data. These general hypotheses concerning procedures for "hand" classification of components are a good starting point that with further detail would improve reproduction of experiments and facilitate comparisons of alternative hypotheses/procedures. Empirical validation could help to resolve questions by supporting or disconfirming the proposed guidelines.