Statistical inference and multiple testing correction in classification-based multi-voxel pattern analysis (MVPA): random permutations and cluster size control

Neuroimage. 2013 Jan 15;65:69-82. doi: 10.1016/j.neuroimage.2012.09.063. Epub 2012 Oct 4.

Abstract

An ever-increasing number of functional magnetic resonance imaging (fMRI) studies are now using information-based multi-voxel pattern analysis (MVPA) techniques to decode mental states. In doing so, they achieve a significantly greater sensitivity compared to when they use univariate frameworks. However, the new brain-decoding methods have also posed new challenges for analysis and statistical inference on the group level. We discuss why the usual procedure of performing t-tests on accuracy maps across subjects in order to produce a group statistic is inappropriate. We propose a solution to this problem for local MVPA approaches, which achieves higher sensitivity than other procedures. Our method uses random permutation tests on the single-subject level, and then combines the results on the group level with a bootstrap method. To preserve the spatial dependency induced by local MVPA methods, we generate a random permutation set and keep it fixed across all locations. This enables us to later apply a cluster size control for the multiple testing problem. More specifically, we explicitly compute the distribution of cluster sizes and use this to determine the p-values for each cluster. Using a volumetric searchlight decoding procedure, we demonstrate the validity and sensitivity of our approach using both simulated and real fMRI data sets. In comparison to the standard t-test procedure implemented in SPM8, our results showed a higher sensitivity. We discuss the theoretical applicability and the practical advantages of our approach, and outline its generalization to other local MVPA methods, such as surface decoding techniques.

MeSH terms

  • Brain Mapping / methods*
  • Humans
  • Image Processing, Computer-Assisted / methods*
  • Magnetic Resonance Imaging / methods*
  • Models, Neurological*
  • Models, Theoretical
  • Sensitivity and Specificity